Changelog

New

• You can share one Interactive HTML block or 3D model on its own link, and the rest of the page stays private
• When an agent you assigned starts the task, you get a notice that they started it

Improved

• In Media studio you can turn Brand style off for one video; the next video uses your saved brand again

Fixed

• On a phone, Back returns you after a slow open, and Forward walks back through the pages you left
• Closing a page in one window keeps your avatar and grid markers on that page in another window
• Chat titles keep words like I'm and it's whole
• The skills tour now points at + (Add context) → Skills

Improved

• Switching workspaces or organizations now shows loading placeholders instead of an empty sidebar and page

Fixed

• Trash is back at the bottom of the sidebar, so you can open it in one click
• On a phone, editing a page or typing a chat message no longer zooms the screen in
• The workspace switcher keeps the last workspace name visible until the next one is ready

Fixed

• On a phone, choosing another workspace in the sidebar now opens it instead of staying where you were

Improved

• Pages you file under Home now sit under the Home row, which opens and closes like a folder
• Every menu now shares one look: rounded, softly shadowed, with the same row size everywhere
• Menus from the sidebar's New button and from each row's “⋯” open beside the sidebar instead of over your pages
• On a phone or tablet, menu rows are taller and easier to tap
• “Sign out” in the account menu is no longer red, because it deletes nothing
• The sidebar's plan meter turns orange near your limit and red at it, instead of staying green
• On a tablet the sidebar is wide enough to read every tool's name, and on a phone your pages get most of the drawer
• Find sources now checks every link before saving it, so dead and blocked pages stay out of your Knowledge Base
• Sources that block reading, such as LinkedIn profiles, now say so and are no longer retried on every refresh

Fixed

• On a phone, starting a comment on a page no longer zooms the screen in
• Export, Version history and ⌘K now take keyboard focus, close with Escape and return you to where you were
• In Arabic, every menu now opens and reads right to left
• The Share panel no longer shrinks to a narrow strip on a phone, and its confirmation shows above it
• A tab's right-click menu closes with Escape and stays on screen, and its “Share…” opens the Share panel instead of publishing a link at once
• Merging a split view keeps the page you were working on in front
• Quickly undoing a tab change no longer brings it back a moment later
• The Content tab, its gallery, the calendar and Home's numbers now follow the workspace on screen, not your first workspace
• Suggested edits in comments now land on the text you selected, even across paragraphs
• A deleted report's saved pages stay in your Knowledge Base, and archiving the last copy asks first
• In full screen, the HTML block's toolbar no longer covers the animation
• A board's “Columns by” menu now closes when you press outside it or Escape, on touch too
• Version history no longer points to a list “on the left” on a phone or in Arabic

New

• Home is now a row at the top of the sidebar, and pages filed under Home show at the top of your pages
• Reports, Content and Knowledge Base sit on top of the sidebar, and every other tool waits under “More”
• Press “Customize” at the bottom of the sidebar to choose which tools stay on top and which wait under More
• Ask the chat to hand a task to an agent or a teammate, for example “Assign this task to Claude”
• Writing one agent's or one teammate's name in a task's Assignee column now assigns them the task
• A task's details now show who requested it, even when it was an agent, and each activity line names who did it
• The ActVox CLI 0.11.0 can assign tasks to agents and @mention agents in comments

Improved

• The sidebar reads calmer: no counts at rest, long titles fade instead of ending in “…”, and one row is lit at a time
• A page shows an icon in the sidebar only when someone picked one
• Recent is one list with “Show more”, instead of groups by date
• The Home row takes your Home page's new name when you rename it
• A tool you hid in Customize before now waits under More instead of disappearing
• Tab names such as Settings, Billing and Knowledge Base now follow your language
• On a phone, a page's header takes one row, and a database's toolbar fits in one more
• On a phone, 👍/👎 sits at the end of a page under “Was this useful?”
• A shared link whose preview can't load now reads “A shared workspace for your team and your AI agents”
• When an organization comes back over its AI connection limit, the extra connections pause, with “Resume” in Settings

Fixed

• Closed security gaps: account details, a workspace's strategy and its connected social accounts can now be read or changed only by people with access
• ⌘K, “Generate with AI”, “Write with AI” and the phone chat now work in the workspace on screen, not in your first workspace
• Deleting a page, database, board or chat now closes its tab in both panes of a split view
• Opening something already open in the other pane now shows it there instead of opening a second copy
• Closing or moving the last tab on the left of a split view no longer leaves an empty pane
• In a split view, the sidebar highlight and the address bar follow the pane you are working in
• A tab opened by a newer version of ActVox no longer breaks the tab bar in an older window
• Switching workspaces or organizations no longer brings back the “Switching…” overlay
• On a phone, the tab sheet no longer sends you back to the workspace you just left
• Undo after deleting the workspace you are in now brings you back to it
• Asking the chat to “explain this in more detail” with a page open now answers instead of failing
• Agents reading several items at once now get shorter previews that fit, instead of an error
• Competitor research no longer saves unrelated social profiles; one it can't confirm is marked unverified and kept out of your Knowledge Base
• A long page title no longer stretches the “Move to” menu off the screen

New

• New accounts start with one question, “Connect my AI”, “Analyze my website” or “Start blank”, and each ends in a finished result
• A workspace started that way shows only the tools you use in the sidebar; “Show all tools” brings back the rest
• Home shows a “Get started” checklist that ticks itself as you go, instead of an automatic tour
• Connected AI agents can now @mention each other, or themselves, in comments, and each one finds the mention in its inbox
• Agents can hand tasks to other agents, or to themselves, in stages, and a person still accepts the last stage
• Owners and admins can stop agents mentioning and assigning each other in Settings → Agent → “Agents working together”
• Each connected AI's page suggests prompts to try, such as “Onboard me to ActVox” and “Review my workspace”
• Write Threads posts and Pinterest Pins in Content, with their own previews, a Pin title and a destination link
• Content cards and draft tabs show each platform's logo
• Add any web page to your Knowledge Base with “Add a link”
• What you ask the chat to remember now shows in Memory under “From chat”, where you can forget it
• The ActVox agent can add and change animated motion blocks on a page; press “Ask AI to change this” on one
• On a phone, Reader shows a page's attachments from the start
• Free accounts get a quiet “Upgrade” button in the top bar, for people who can buy
• Agents can read, archive and restore generated artifacts through the API and MCP
• actvox.com's header adds “Create free workspace”, and a plan chosen on the pricing page opens Billing with that plan selected

Improved

• Rename, share or make an AI assignable from its own page, under Name, Mentions and tasks, Who can use it and Inbox
• On a phone, nothing floats over the bottom of a page: “Open tabs” and “My tasks” sit at the top of the navigator, and chat opens from the top bar
• On a phone, a failed save shows “Save failed · Retry” in the top bar, where it stays in view
• Home's greeting now reads “Where your team and your AI agents keep their work.”
• The prompt that connects an AI, and the prompts it suggests, are written in your language: English, Russian, Ukrainian or Arabic
• Agent cards in Connectors show the agent's own name, its permissions in words and the last page it saved
• Knowledge Base cards read like “Ready · 12 sections”, with plain statuses instead of chunk and token counts
• Adapt now shows that one run covers up to three platforms, instead of quietly dropping the rest
• Sign-up asks only for your email and a code, offers an email link, and no longer asks for a username
• Sign-up opens in the language you were reading actvox.com in
• On actvox.com, “Create your free workspace” on the Agents page starts your new account on connecting your AI
• actvox.com's header now reads Agents, Use Cases, Pricing and Help, and Roles moved into Use Cases
• The pricing page no longer calls Pro “Most popular” or lists features that every plan already has
• The Terms now count credits the way the product shows them: 1,000 ActVox Credits cost $1
• Upgrade and billing messages no longer promise plans or limits the product doesn't have
• The sidebar's plan meter offers an upgrade only for limits that apply, and only to people who can buy
• When you decline analytics, ActVox's servers stop sending your usage events too
• Sidebar tooltips and connector instructions use one name for each thing and point to places that exist

Fixed

• Asking the chat to “remember that…” now saves it, instead of getting an unrelated answer
• When today's free chat messages run out, chat says when they come back and offers “Buy credits”, instead of an error
• Creating a Project from a report at your workspace limit no longer crashes the app
• At an organization's workspace limit, admins get “Upgrade” and members are told to ask an admin
• Buy credits and Upgrade prompts open Billing on the right section, even before you finish your profile
• Plan and credit buttons are enabled only for people who can actually buy for the account
• Checkboxes in Settings now tick when you click the box itself, not only its label
• A new draft's title shows in the studio header instead of “Untitled draft”
• Knowledge sources no longer turn Failed when a monthly refresh can't reach them; they keep what they had and retry
• Search results show a page's words instead of pieces of its code
• “Ask AI to change this” on a motion block now changes the block, and page edits that only mention animation stay page edits
• New visitors who choose “Create your free workspace” reach sign-up, not sign-in, and keep the plan they picked
• An invitation to a page opens that page after sign-up, instead of a new workspace
• On a phone, the comment box stays in view in long threads, and typing a comment no longer zooms the page
• A carousel's photo count stays in its corner on right-to-left pages
• Sign-in email links again work only on the device that asked for them
• “Read the guide” on Reports and Calendar opens the right Help Center section

New

• Comments on selected text now sit in the margin beside the words they're about, and the text moves aside to make room, even with the sidebar open
• When there's no room beside the text, a comment opens in a small window at its text, and on a phone from the bottom of the screen
• React to any comment with an emoji: reactions show as chips under it, and pressing yours again removes it
• Hover a comment for its actions: “Add reaction”, “Resolve” on a thread's first comment, and “Edit”, “Delete” or “Copy link” under ⋯; on a phone they stay in view, clear of the comment's time
• Comment on a whole block with “Comment on this block” beside it, or press ⌘⇧M with nothing selected
• On a phone, code blocks wrap long lines, with “Wrap” and “Scroll” to choose between; your phone remembers the choice
• On a phone, a table wider than the screen gets “Open full screen”, as do HTML blocks on shared pages; Back closes it
• Paste one prompt into Claude Code, Codex, Hermes or OpenClaw and it connects itself; you only sign in and approve in your browser
• Copy that prompt from the new “Connect your AI” box on Home, from an agent's card in Connectors, or from actvox.com

Improved

• Resolved threads and comments on deleted text now wait behind one “n more comments” list, which closes when you open a thread; Escape closes the panel on top
• Comments on a page or a database row fold into one row that shows who took part, how many there are, and who wrote last
• Page comments start open when there are one or two, one mentions you, or something is new; otherwise your choice is remembered
• Long threads show the first comment and the newest reply behind “Show n earlier replies”, and long comments fold behind “Show more”
• Comments show simple formatting such as lists, bold, code and links, and times such as “2h ago”
• An agent's status at the end of its comment now shows as a small chip, with its details in a quieter line
• An agent you @mention in a comment is named in the text instead of in a second row of chips
• The workspace downloads about 90 KB less code on a first visit, so it opens a little sooner on phones
• Pages opened from a link or after a reload appear sooner, by about 0.7 seconds on a phone
• On a phone, a page never scrolls sideways: long words, links, headings and titles wrap, on shared pages too
• Claude Code's setup command now connects ActVox in every project, not only the folder you ran it in
• Every page on actvox.com now opens with the same header style
• The ActVox logo at the top and bottom of actvox.com no longer sits on a white tile and takes the menu bar's text color
• actvox.com/llms.txt now explains why ActVox exists and walks an AI agent through working with it, step by step
• actvox.com's robots.txt now names AI search, assistant and training crawlers, with the same access as before

Fixed

• “Reply” and “Resolve” now show on page comments without first pressing “Comment”
• Comment avatars no longer show punctuation in their initials: “Claude (MCP)” now reads CM, not C(
• The Codex command-line app can now sign in to ActVox instead of being turned away at the consent page
• Setting up Codex no longer asks you to sign in a second time after it has already signed you in
• Section headings on actvox.com are back in their monospaced typeface, which they lost in 3.28.0
• The Use cases page title on actvox.com is no longer in all capitals
• Right-to-left pages, such as Arabic ones, no longer shift sideways while you edit them on a phone
• A gallery set to “Full width” now spans the page on its public link too, edge to edge on a phone
• A carousel's photo count reads “1 / 8”, not “8 / 1”, on right-to-left pages

New

• Press “Record a voice note” in Capture and ActVox writes out what you said, keeping the recording with the note
• Capture works without a connection: notes, voice notes included, wait on your device and become pages once you're back online
• The end of a page now lists the pages that link to it under “Linked from”, newest first
• Pages the ActVox agent writes now end with “Sources”: the knowledge-base files and web pages it cited, named by their site
• Link previews of a publicly shared page now show a picture of the page
• Attach up to three photos to a chat message and the agent looks at them, including any text they show
• Each chat message now says which attachments were sent to the agent, and why any were left out
• “Ask the agent” in Capture can now send photos too
• See every AI app and agent key you connected under Your agents in Skills and Connectors → Connectors
• Switch Your agents between Mine and All my workspaces, and pause or revoke a connection right from its card
• Replay the tour of the screen you're on with “Take the tour” in the account menu or the command palette
• When someone is arranging a grid on a page, everyone else sees who; joining in takes “Edit layout anyway”

Improved

• AI agent cards in Connectors now say “Connected”, or how many of your agents are connected, instead of a test level
• Settings → Connections keeps everyone's connections for owners and admins, with a summary row and one card per connection
• Each onboarding tour now starts by itself only once per person, in one tab, whatever workspace or device you use
• The “Keyboard shortcuts” tip now shows by itself only once per person, and not on phones; open it anytime from ⌘K or the account menu
• actvox.com now presents ActVox as a shared workspace for teams and their AI agents, with a prompt to connect yours and “Talk to the founder”
• Monospaced text such as numbers and dates now uses the same typeface in Russian and Ukrainian as in English

Fixed

• On a phone, Back now returns to the tab you opened something from instead of leaving the workspace
• “Last call” now counts the ActVox CLI, so agents such as Hermes and OpenClaw no longer show a dash while active
• An expired connection now reads “Expired — reconnect” instead of Active, and Your agents offers “Reconnect”
• A connection that expired while paused can now be reconnected instead of only revoked
• When your browser blocks copying, ActVox now says it couldn't copy instead of saying the change failed
• Text a connected agent replaces on a page keeps its spaces and line breaks, and replaced blocks stay editable by later agent edits
• Zooming a picture in an interactive HTML block now stops before the picture turns blurry
• Connected agents listing your content now get every post's text, no longer crowded out by pictures stored inside posts
• Adding ActVox to an iPhone or iPad home screen now shows the ActVox icon at every size

New

• A new notifications bell in the top bar collects your mentions, replies, tasks, reviews and approvals under “For you”
• Notifications are written in your own language, whoever caused them, and open right at the comment, task or page
• Mentions, replies, assignments, reviews and approvals also pop up for a few seconds while you work
• Choose which notifications reach your bell, and turn pop-ups off, in Settings → Notifications
• Every change to a page's name, icon, cover or place, and to tables, rows, whiteboards and folders, is now recorded with who made it: a person, an outside agent or ActVox's agent
• Choose History from an item's menu in the sidebar to see each change as before and after, filtered by people or agents
• Restore an earlier version from History; the restore is recorded as a new change, so you can undo it too. A page's text keeps its own versions under More actions → History
• Anyone who can edit a workspace can now connect their own agent, private to them until an admin shares it
• When you connect an app, ActVox asks “Which agent is this?” so you can pick one of your agents or name a new one with its own @handle
• Owners and admins can share an agent with the whole workspace or make it private again
• Settings → Connections lists everything you connected across all your workspaces, and owners and admins can see every member's connections and the whole organization's
• Every agent shows how sure ActVox is which app it is: Verified app, Callback verified or Declared
• Muse, Gemini, Gemini CLI and Kimi Code can now connect to ActVox, each with its own setup steps in Skills and Connectors
• Grok Build, Grok Bot and Dots can join as your own agents through the ActVox CLI or an agent key
• Apps that can't use the ActVox sign-in, such as Z Code, can connect over MCP with a workspace API key
• Hermes and OpenClaw can connect themselves from a prompt you paste in, stopping only for your approval in the browser
• Your Telegram agent can suggest changes to an existing page, and its reply opens the suggestion for you to accept or reject
• Install ActVox as an app on your computer or phone with “Install app” in the account menu
• Settings → Apps and a “Take ActVox with you” card on Home show how to install, with a QR code for your phone
• Join the desktop app waitlist in Settings → Apps and we'll email you when it's out
• Capture a note, link, photos or files from Home or the installed app, then save it as a page or hand it to the agent
• On Android, share from any app straight into ActVox once it's installed
• Give a whiteboard its own background colour; collaborators, shared pages and previews see it too
• Keep up to 50 named characters for motion blocks in Brand Visuals → Characters, or save one from a scene, and pick them in any character scene
• A block keeps the look a saved character had when it was made, until you press “Update to latest look”

Improved

• Empty trash and Delete forever now wait 30 days, so you can still restore or undo; owners and admins can choose Delete now
• An agent never does more than the person who connected it: an editor's agent can't share, publish or change workspace settings
• On plans with an AI connector limit, the limit now counts every member's connectors in the organization, and an app a teammate already connected adds none
• Reconnecting or revoking one connection no longer removes its agent, which keeps its @handle, history and sharing
• The @mention and assignee pickers show whose agent each one is and mark your private ones
• An agent key made on a connector tab gives you your own assignable agent at once, and Settings → API Keys asks which app a key is for
• Connector tabs show each app's own logo instead of a plug
• Each member can pick the language the agent answers them in, under Settings → Agent → Your language
• Only owners and admins can change workspace-wide agent settings such as the default language, memory recall and auto-adapt
• Hover the ActVox logo in the top bar and click it to open or close the sidebar
• Videos made with the Fast model can now be 15 seconds long and come with their own sound
• Pop-up messages follow your theme, and sit on the correct side in Arabic
• Notices about paused auto-adapt, skipped adapts, an ending plan or paused indexing now say what happened, and the indexing and plan notices go to admins
• Fonts now load from ActVox itself instead of an outside font service
• ActVox skills 3.38 teach agents to use saved characters and explain what an agent may do within its person's role
• ActVox CLI 0.10.1 fills in the agent's name on the consent page, also when you sign in with a code on another device
• Skipping the onboarding tour now hides it in every workspace and on every device

Fixed

• Edit Image opens your picture again instead of an empty box, and shows a single Close button
• Re-opening a saved image edit keeps its full size and layers, even after a week
• Edited images are saved at the picture's full resolution, and a failed load or save now tells you
• ChatGPT and other apps that make several calls at once no longer lose answers from ActVox
• An organization admin who is demoted no longer keeps admin rights in ActVox
• Your theme and language now follow your account, so the next person on a shared computer doesn't inherit yours
• Task due reminders now reach every task in every table and go to the people it is assigned to
• Content sent for review now notifies every member of the workspace, not only the person who asked
• An approval notification opens the approval for review instead of the workspace home
• One organization's billing alerts no longer show in another organization's workspaces
• Agents hand back their open tasks when they are disconnected, their key expires, or their person becomes a viewer or leaves
• Connector tabs show only your own connections and keys, and open for members again instead of failing
• The built-in workspace agent can no longer be revoked from a connector tab
• The Recent list in the sidebar no longer shows other members' agent tasks
• In dark mode, whiteboard stickers, emoji and pictures keep their true colours
• An interactive block pasted into another workspace keeps its files
• Typing into a selected table cell no longer scrambles the letters
• Typing a new page's title no longer adds to the word “Untitled”
• Screen readers now name every emoji and picture in the page icon picker
• The chat box's placeholder, buttons and menus now appear in Russian, Ukrainian and Arabic instead of staying in English
• A notification about a revoked invitation now opens Manage organization on its Members page

New

• Add a Characters scene to a motion block: pick up to two cartoon characters, dress them and write what they say
• Characters can be a person, a dog or one of our mascots, with hats, glasses, a whistle, a clipboard and more
• Connected agents can create character scenes too
• Create a whiteboard inside a page from the page's menu in the sidebar

Improved

• ActVox skills 3.37 teach agents how to write character scenes

Fixed

• SVG pictures pasted or dropped on a whiteboard now save and show for everyone
• A picture the browser can't open is removed with a note instead of leaving an empty placeholder
• A whiteboard's right-click menu no longer hides under the side panel

Improved

• If a tab has been open since before this release, reload it before editing so new block settings aren't lost

Fixed

• Images you place on a whiteboard are now saved and show up for collaborators, on shared pages and in exports
• Undo now works right after pressing Enter or inserting a block
• Inserting a gallery, or switching it to even rows, now undoes in one step
• After using a callout, gallery, image or Mermaid control you can keep typing, or undo, without clicking back into the page
• In right-to-left languages, a callout's colour button no longer covers its icon
• Pressing Undo can no longer empty a page that was just repaired after a sync problem
• On whiteboards, the A key now draws the same straight arrow the toolbar shows
• Whiteboards with pictures keep their preview on shared pages and in exports, even with several boards on one page
• Reconnecting an AI app you already had connected no longer fails when your plan's connector limit is reached

New

• The Pen on a whiteboard now opens a Marker, a Highlighter and Washi tape, each with its own colours
• Sticky notes come in ten colours and can show the name of the person who wrote them
• Stamp a thumbs up, a star, a heart or your initials on a whiteboard, or add an emoji from the Stickers tab
• Whiteboard connectors can be straight, curved or elbowed, and templates show a preview before you add them
• Give a callout its own emoji or image and one of nine colours
• Type : and a couple of letters in any text to insert an emoji
• Show a Mermaid diagram as its code, as the picture, or both side by side
• Replace an image and keep its size and caption, and download any image you can see
• Images, galleries, embeds and videos can sit flush against each other, and a video can go wider than the text
• Galleries can lay photos out in even rows that keep each photo's shape, with a 0, 4 or 8 px gap
• Connected agents check their mentions and tasks in all your workspaces in one step, skipping those where they aren't assignable
• Connected agents can acknowledge, claim and hand in their tasks directly, and know their own @handle in each workspace
• Agents can read a page as markdown, which takes about a quarter of the space the old format did

Improved

• The Eraser now sits with the Pen's drawing tools instead of under More tools
• On images and galleries with no spacing, the controls float over the photo instead of taking a row
• Agents can now see which agents a comment mentions, not only which people
• ActVox skills 3.36 teach agents the new block options, reading pages as markdown and taking a task from the inbox

Fixed

• Assigning a task in the Assignee column now gives the agent the row's text as its brief, not just the title
• An agent you have just made assignable picks up that workspace on its next check instead of skipping it
• Each workspace an agent checks in one step now shows that check on its Connections card
• The inbox routine in ActVox skills now uses commands the CLI accepts, so scheduled checks can reply and acknowledge
• The prompt to schedule inbox checks no longer appears for a paused connection
• After renaming an agent, removing its old @handle from a comment now removes the mention
• An admin who cannot rename an agent no longer sees Rename
• “Show system databases” and Memory find the right databases in very large workspaces and after old copies were trashed
• Adding an AI app your organization already uses to another workspace no longer runs into the connector limit
• Retrying a connection no longer loses the note about which workspaces were left out
• Pasting a callout no longer leaves its icon behind as stray text
• On shared pages, a callout's image icon no longer opens as a full-size photo
• In the Motion scene panel, a read-only image scene can no longer be deleted, and ⌘S and Esc work while a picker is focused
• Motion blocks from agents are no longer refused for text the published limits allowed
• An HTML block you are editing no longer overwrites a newer version an agent saved meanwhile
• Agents allowed only to read databases can now open each block of a row they can read
• A long page read by an agent now always stays within the size the agent asked for

New

• Insert a Motion block from the / menu and edit its scenes: titles, lines, a big number, lists, steps, a quote, a countdown or a breathing timer
• Claude, ChatGPT and other connected agents can create motion blocks on your pages and change them later
• “Make assignable” now offers Claude.ai and Grok, and preselects the app an agent connected from
• Agents can read one block of a page by its id, including a large interactive block in pieces

Improved

• Images and files on your pages now load from api.actvox.com
• Connected AI apps now show their verification as a lasting status that cannot expire, with a tooltip explaining what it means

Fixed

• Grok's connector no longer shows “Verified read” next to a write it has already verified
• Claude Code and other agents can read a long page in pieces instead of getting a file too large to open

New

• Agents can publish an interactive block of up to 16 MB, such as a detailed 3D scene, without shrinking it first
• An assignable agent's connector page gives you a ready prompt to schedule, so the agent checks its mentions and tasks on its own
• Connected agents now know to check their ActVox mentions and tasks when you ask them to work in ActVox
• The Assignee column lists your assignable agents next to people, so you can hand a task to an agent there
• A task's Assignee cell shows the owner and where the work stands, from Assigned to Done, and opens the details on click
• You can rename an agent and give it its own @handle from its connector page
• CLI agents can read their mentions and tasks across all their workspaces with actvox inbox (ActVox CLI 0.10.0)
• Pages can play motion blocks, short animations with play, pause and scrubbing; creating them arrives in a later release

Improved

• Dropdowns and date, time and colour fields now use ActVox's own pickers instead of the browser's
• Memory opens in its own window, and Knowledge Base and Memory now sit under Files & Media in the sidebar
• You can hide Knowledge Base and Memory from the sidebar in Customize
• Publishing a large interactive block is faster
• Pressing Escape in an open menu now closes only the menu, not the panel behind it
• People show their profile picture in pickers and on cards, and agents show their app's logo
• Assigning now happens in the Assignee column alone; the separate Assignment section under a row's properties is gone
• An agent shows “Waiting for first check-in” until it first reads its inbox, instead of “Unverified”
• The connection consent page shows how many AI connectors each workspace on a capped plan has used

Fixed

• Knowledge Base and Memory no longer open and highlight the same window
• “Show system databases” now explains what it reveals, and says so when a workspace has none
• An agent that isn't assignable yet is now pointed to the right screen to fix it
• Board, gallery and calendar cards show a row's icon instead of its code
• Reconnecting Grok or another AI app no longer counts as a new connector or blocks the connection
• When a plan's connector limit stops a connection, the consent page names the workspace and links to upgrade it

New

• One block can go wider than the text column while the rest of the page keeps its reading width
• Shared pages built by an AI agent say so under the title, and you can hide that line from the Share panel
• The Share panel shows how many times a public link was opened, and by how many people, over the last 30 days
• Interactive blocks built by agents can use textures, 3D models and fonts stored in your workspace files
• Agents can send an interactive block's data as a file and update its HTML and data in separate, smaller steps
• A new Agents page and an agents section in llms.txt explain how to connect Claude, ChatGPT, Grok and Codex
• An admin can make a connected ChatGPT, Claude or Grok an agent your team can @mention in comments and assign tasks to

Improved

• The Personal plan now leads with what it includes: it runs on the AI subscription you already have
• The site now says agents can read and write in your workspace, not only read it
• ActVox Research is now simply ActVox across the site, shared pages and help
• People now appear by name in the assignee picker, @mentions and comments, with their email underneath, instead of a company name or a bare email
• Help now gives the setup steps for connecting Codex, Hermes and OpenClaw
• The agents section of llms.txt now lists which clients have documented setup and the size limit for a block's files

Fixed

• Shared pages and PDF or PNG exports now show tables of contents, formulas and whiteboards instead of grey boxes
• A toggle you closed now stays closed on shared pages and in exports
• Signing in or signing up to duplicate a shared page brings you back to finish the copy
• “Create your own workspace” on a shared page now opens sign-up instead of sign-in
• Fixed two Russian typos on shared pages and the “Operational System” typo in the homepage headline
• An agent's page edit is no longer refused just because someone has the page open
• Pages with long markdown tables no longer hit the page size limit
• The Person picker, assignments and @mentions now list the same people in an organization workspace
• Comments and mentions now work in organization workspaces that were created without an owner
• Interactive blocks are now checked the way a browser reads them, so hidden markup can no longer slip past the safety check
• Interactive blocks no longer disappear in Safari a moment after they load
• Copying an interactive block to another page now keeps the files it uses
• The pricing page now names every plan, including Grow and Enterprise

New

• A connected agent that cannot reach file storage directly can send a 3D model, image or HTML file in small pieces, up to 5 MB.

Improved

• ActVox skills 3.27.0 teach connected agents to send a file in pieces when one call is too small for it.

Fixed

• After editing an HTML block, Preview shows it exactly as it renders again and keeps its current slide.

Improved

• Billing shows one available balance with what it is made of, and today's spending and free chat turns on one line.
• Recent activity describes plan changes and renewals in plain words, and a renewal that adds no credits no longer shows +0.
• Usage and activity now comes before Workspace services on the Billing page.

Fixed

• In a personal workspace, a card's Person field offers only you, not the members of your organizations.
• Workspace services no longer stays on Calculating workspace usage; a workspace not yet counted says so.
• A Personal plan without monthly credits is no longer told to choose a plan.

New

• Database tables have a column menu with freeze, wrap and twenty footer calculations, plus Filter, Sort and Search in the view bar.
• Open a table row beside the table in a side peek, drag rows to reorder them, and change many selected rows at once.
• Page grids are available to everyone: put text, images and HTML blocks side by side, and agents can build them too.
• On a phone you can comment while reading, review an agent's suggested changes, and reorder a page's blocks.
• Accept an agent's proposed page edit straight from its preview link, with the changed blocks outlined.
• When an agent working in Telegram needs your approval, it sends a link that opens Approve and Reject in the app.
• A new Grow plan for organizations at $49.99 per seat, with 75,000 ActVox Credits and 5,000 blocks per seat.
• Enterprise is available on request, with workspace, block and credit limits set by agreement.
• A Personal plan at $19.99 a month for one person, with AI-generation credits bought separately.
• Promotion codes can be entered at checkout for subscriptions and credit packs.
• Billing shows the pages, databases, whiteboards, knowledge sources and stored files your workspace uses.
• Generated images and carousels can stamp your logo as uploaded or in a transparent version from Brand DNA.
• Your plan is shown next to your name in the account menu and the organization switcher.
• A page that is already shared shows Copy public link or Copy organization link right in its preview.

Improved

• Every paid plan now includes ten workspaces; Free includes five.
• Pricing separates personal plans from organization plans priced per seat, and describes content in shared blocks across your workspaces.
• The workspace loads about a third less code on a first visit, and pages open faster from database cards, mentions and Home.
• On phones a page's text appears before the editor finishes loading, and your reading position survives the switch.
• Opening or closing a split view no longer reloads every open page.
• Checkout shows team prices per seat and the ActVox logo on every item.
• Page invitation emails are limited per hour and per day, and the share panel says which limit you reached.
• Home's Connectors widget puts Claude, ChatGPT and Grok first and shows whether you have connected each one to this workspace.
• A link to a comment opens the page at that comment, from a preview as well as the editor.
• Workspace, report and invitation pages ask search engines not to index them.
• ActVox skills 3.26.0 and CLI 0.9.0 let a connected agent find your instructions and design pages without creating them.

Fixed

• A tab order you just dragged is kept after a reload.
• The sidebar no longer shows Calculating block usage or Block usage needs recovery to every account.
• Going back from an approval link takes one press, and a proposal link that was already handled says so.
• Signing out on a shared device now clears your data from the browser, so the next person starts clean.
• The welcome tour no longer covers the workspace, and the welcome window no longer returns after every sign-out.
• Open in a page preview works after reloading on a chat tab.
• A row opened from a database inside a page no longer closes itself, and formulas on a row's page find their columns.
• Escape closes popovers such as Filter and Sort, and a new row in a sorted table no longer types into another row.
• Column lines sit on the correct edge in Arabic.
• Artifacts and Knowledge Base toolbars wrap in narrow panes.
• Share link previews for interactive HTML pages show the page's own image.
• Frameless HTML blocks keep their full height while you edit them.
• Chats left open for a long time renew their connection cleanly instead of timing out.
• A bank payment that fails after checkout no longer leaves the purchase stuck, so you can try again.
• Organization billing stays on the organization, and monthly allowances show the right number of credits.
• Earlier subscriptions stay manageable after checkout moved to Stripe.

New

• The Trash has an Empty trash button that deletes everything in it for good and shows how many items are gone while it works.
• A guided start for outside agents: the Connectors prompt has your agent read the workspace, suggest one small first step and wait for your yes.

Improved

• The Connectors tab now starts with a short connection check: the agent names the workspace and its permissions and changes nothing.
• The first prompt for a connected agent names the workspace it was copied from, so an agent with several workspaces works in the right one.
• ActVox skills 3.22.0 and CLI 0.8.1 give agents clearer guides: size limits for HTML blocks, 3D models versus Three.js scenes, and updating the CLI to an exact version.

Fixed

• Delete forever can no longer take live pages, boards or databases along with a deleted parent; what is still in use is kept and moved up a level.
• Deleting a large database forever now works past 1,000 rows and clears its rows' history.
• A page nested too deeply to save now says so and keeps your edits, instead of quietly not saving.
• Fullscreen now works for HTML slideshows on pages, in the artifact view and on shared pages.
• Copy page and View as Markdown now include the text of interactive HTML blocks instead of only the page title.
• A connected agent's read-only first check no longer creates an instructions page in a new workspace.
• Page invitation emails name the person who invited you instead of showing a bare email address in the subject, and a reply goes to that person.
• Accepting a page invitation no longer fails with a different-email error when you are signed in with the invited, verified address.

New

• A link to a page in an agent's chat answer opens a quick preview over the chat, scrolled to the part the link points at; Back returns you to the conversation.
• A link to an edit the agent suggested opens the page right on that edit, with Accept all and Reject all ready to use.
• Interactive HTML blocks now run on iPhone, iPad and Firefox, on shared pages too, instead of showing only their text.
• Images, videos, PDFs and HTML artifacts in Files & Media and Artifacts open in a full-screen viewer you can step through with the arrow keys.
• Files & Media and Artifacts keep loading past a few hundred items instead of quietly stopping, and search looks through all of them.

Improved

• Full width now widens everything on a page, text included; Standard width keeps text in a comfortable column and is the default.
• A big database in the sidebar shows its first 8 rows and a Show more row instead of a long block of names.
• Agents connected through ActVox's MCP tools now see exactly which fields each tool accepts and get clearer errors when a request is wrong.

Fixed

• Accepting an agent's suggested edit no longer fails when the agent changed a word or sentence inside a paragraph.
• Saving in an HTML block's Edit mode no longer reloads the artifact and resets it, such as a deck jumping back to its first slide.
• Suggestion chips that point at a section of a page now open that section.
• Removing or replacing an image on a page no longer breaks the same image in a copy shared into another workspace.
• The sidebar, Trash and page counter load reliably in very large workspaces, and the nightly cleanup of old reports runs again.
• Confirm dialogs no longer push their buttons past the edge, and the Share dialog's view and edit choices use a proper menu.
• Tables with resized columns, videos and embedded players now fit the page width instead of freezing or letterboxing.
• Long values on board and gallery cards wrap instead of being cut off mid-word.
• Three agent tools that always failed — agent notes, accepting a memory proposal with changes, and patching a whiteboard — now work.

New

• Interactive HTML blocks can pin Three.js's BufferGeometryUtils (actvox:lib/three-buffer-geometry-utils@0.183.0, loaded after three@0.183.0) — geometry merging without bundling a second copy of Three.
• Agents rendering a block can read applicationStatus — ok, blocked or unknown — to know whether the block's script actually ran, not just that a screenshot exists.
• An agent's receipt for a page it created or edited now links straight to that page — to the proposal or the first changed block when there is one — and Telegram results link the page the run wrote.

Improved

• While an organization switch resolves, the workspace keeps its shape — the sidebar column stays with a single "Loading workspace…" line and the switcher says "Switching workspace…" — instead of collapsing to a lone spinner.

Fixed

• A tab whose organization was switched elsewhere — another tab or window — no longer keeps showing the previous organization's workspaces under the wrong tenant: the workspace reloads itself within about ten seconds.
• The sidebar page meter no longer breaks its count across three lines next to the "Upgrade the organization" action.

Improved

• A Page you leave and come back to reopens instantly, at the same scroll position and caret, instead of loading from scratch.
• Pages are prefetched from Recents, Pinned, the ⌘K palette and shared links, so they open faster.
• The sidebar page meter counts past the Free limit ("59 of 25" rather than a stuck "26 of 25") and, on an organization's workspace, asks you to upgrade the organization — the plan that owns those pages.

Fixed

• Switching organizations no longer changes the sign-in claim while the previous organization's workspace is still on screen, and no longer reloads the page mid-switch.

Fixed

• Switching organizations no longer opens the previous organization's workspace before the one you picked, and no longer leaves it on screen when the new organization has no workspace yet.
• An organization switch makes fewer page reloads and opens that organization's active workspace, not just its first one.

New

• Edit mode on an HTML block can change text inside SVG drawings — floor-plan labels, chart captions, dimensions — and Save keeps it, including text a script draws.
• Agents see which fields every page edit and every workspace action takes before they call, and a refused call now says exactly what was wrong.
• The change feed tells agents who changed each page, row or database — a person or a connected agent — and which blocks or cells moved.

Improved

• Each pending invitation shows its email's real status — sending, sent or not sent — and Resend sends it again.
• The artifact page's top bar shows only the back link and the title.
• Database columns: a duplicated option label is reported, person cells are checked, an import refusal lists the exact rows, and clean conversions no longer warn.
• An agent batch that would change more than one page is refused before it starts, so a failure can no longer leave pages half-updated.

Fixed

• The Share dialog no longer says an invitation email was sent before it went out.
• Accepting a page invitation no longer fails with a bare Server Error: a brand-new account can accept, and a mismatched email address gets a clear message.
• Billing: a referral that stops earning stops paying its affiliate, an out-of-order Stripe event can no longer undo a newer change, a won chargeback restores the credits it took, and an upgrade paid from your credit balance grants its allowance.

New

• Interactive HTML blocks can show live rows from a workspace database and redraw when the data changes, without reloading.
• Only code an agent wrote through ActVox receives live data; if someone edits it later, the block shows its saved data and says live data is paused.
• Interactive HTML blocks can use d3, ECharts and three.js, so agents can build real charts and 3D scenes.
• Agents can render an HTML block the way ActVox shows it and get back a picture, its size and any errors it logged.
• Agents can place a large HTML widget from an uploaded file instead of pasting it into one request.
• ActVox CLI 0.8.0 adds actvox render block and actvox render get, and uploads accept HTML files.

Improved

• Pointing at a page in the sidebar or a tab starts loading it, so it opens almost as soon as you click.
• Opening a page for the first time takes about half as long as before.
• Links that point to a block, from chat, agents or an HTML block's full view, now open the page at that block on desktop too.
• When an HTML block is rejected, agents see every problem with its line number at once instead of one per attempt.
• Page reads list a fingerprint for each HTML widget, so agents can confirm it was saved exactly as sent.

Fixed

• Characters like · and non-Latin text no longer come back escaped when agents read pages through the ActVox skills.

New

• On a phone, pages open in a Reader: scrolling never moves the cursor or a block, Contents jumps to any heading, and Edit switches to editing.
• The Reader remembers where you stopped on each page, and a link to a block opens the page at that block.
• Every connector setup page has Copy page, which copies the whole setup as Markdown for your AI tools, with plain text and View as Markdown in its menu.
• Commands, prompts, config and URLs on setup pages sit in labelled blocks with a one-click copy button, and terminal commands are colour-coded.

Improved

• Switching back to any of your last five open pages is instant, with no reload.
• The page menu's copy control is now Copy page, with the same Markdown, plain text and View as Markdown options.
• Setup prompts use the reading font, and commands mentioned in setup text show as code.

Fixed

• Connector pages fit phones: the intro no longer squeezes beside the header, and the chat button no longer covers the last row.
• A copy button no longer says Copied after you switch to another package manager's command.
• Hosted connector setup no longer tells you to send the first prompt twice.
• People who can only view a page no longer see the block drag handle.

New

• A new ActVox CLI card in Skills and Connectors sets up any agent or script that can run terminal commands, with a copyable instruction for the agent.
• Choose the page where agent-created pages land by default, under Settings › Agent › Agent pages.
• Hermes and OpenClaw can now connect to your workspace, with copy-paste terminal steps and a prompt a local agent can run for you.

Improved

• Workspaces with more than 500 pages show a Load more pages row in the sidebar instead of hiding older pages.
• Setup prompts and config snippets on connector pages wrap instead of scrolling sideways.

Fixed

• OpenClaw's remove step now names commands that exist.

Fixed

• Connected agents creating a folder are now offered the sidebar's real sections, so folders for uploads and AI-made files no longer fail.
• Memory recall no longer ranks a less relevant fact above the answer just because it was saved a minute later.

New

• HTML blocks can drop their frame and take their content's height, or become a page of their own with Turn into page.
• Every code block now has a Copy code button, in chat, previews, report blog posts, shared pages and Help Center articles.
• Readers of a shared page can open any image full-size and step through the page's other images.
• An empty column prompts you to type / for blocks, and the columns menu can reset widths or unwrap the columns.
• Agents can upload images, documents, audio, video and other files to your workspace and place them on pages.
• Agents can list, read, create and update your workspace skills, and import them from GitHub or a ZIP file.
• Pages and databases can carry notes for agents, including rules on what they may edit, create or archive, and ActVox enforces them.
• Agents can change a page's icon and cover, including a cover from a photo they uploaded.
• ActVox CLI 0.6.0 uploads any file with actvox upload and can list, archive and restore uploads.
• Agents can add blocks inside columns, toggles, lists, quotes and table cells, and add HTML or 3D blocks to existing pages.

Improved

• Pages read in a centred column about 75 characters wide, with larger body text; wide blocks can still use the full width.
• Choose Wide blocks: text width in a page's More actions menu to keep every block inside the text column.
• Chat tabs read in a centred column with larger text and calmer headings.
• Labels across the workspace are plain sentence-case text instead of small uppercase code-style type.
• The sidebar, header and tabs are quieter so the page stands out, and inactive tabs no longer look faded.
• One Connected accounts icon in the header replaces the platform letters and shows when an account needs reconnecting.
• Settings tabs, headings and descriptions are larger and easier to read, and text boxes match the other fields.
• Tables no longer push the page down when you hover them, and table columns can't be squeezed to a sliver.
• Two columns stay side by side on narrower screens, and many columns stack before they get too narrow to read.
• Shared pages show an HTML block as one clean frame at its own size.
• The Refer & Earn link has left the sidebar while the programme is redesigned.
• Agents get error messages that name the field or block to fix when a page request is invalid.

Fixed

• Assistant replies, callouts and code blocks are readable again in dark mode, and diagrams now follow the dark theme.
• Interactive HTML on a shared page follows the page's theme instead of the reader's computer setting.
• On phones, report score tiles no longer overflow, and the chat button no longer covers the end of a report.
• Finished reports no longer show a running timer, and an empty Sources tab now says why.
• A partial report now lists its missing sections by name, in your language.
• Very low engagement rates now show as <0.1% instead of 0%, which looked like no engagement at all.
• TikTok and YouTube names and confidence badges stay visible on narrow social cards.
• Code in a report's Blog tab keeps its line breaks, and copy buttons now appear on tablets.
• Bold, italic and links now work inside table cells.
• Knowledge base sources are now strictly limited to their own workspace.
• An agent that names a block that does not exist gets a clear error and never edits another block.
• Agents can add image blocks again, and a second file upload no longer fails while another is in progress.
• Connected agents can send feedback about ActVox again.
• Settings and the Connectors tab no longer show English text in other languages.

New

• Comments are now threads: reply under any comment, resolve the discussion when it is settled and reopen it later.
• Connected agents can read comment threads, reply, resolve them and mention teammates, who get notified.
• Comments written by an agent carry an Agent badge, so you always know who said what.
• 3D model blocks can be edited in place: alt text, a caption, auto-rotate, camera controls, height, poster and AR file.
• Agents can upload 3D models from a file or a public link and add, change or remove 3D model blocks.
• Agents can make a page full width and set the width, alignment and height of images, embeds, files and HTML blocks.
• Your connected agents now learn when their ActVox skill or CLI is out of date and ask you before upgrading.
• Agents can total a database by up to three columns at once, such as revenue by region and month.

Improved

• Embeds, HTML, interactive HTML and 3D model blocks now all resize from their bottom edge.
• You can now delete your own inline comments and replies; a comment that has replies is resolved instead.
• Rollup columns now show their values in database tables, and a rollup of a select shows the option's label.
• Relation cells show the linked row's title, and long lists collapse to three chips and a +N count.
• Rollups now add up to 1,000 linked rows instead of 200, and a cell that reached the limit is marked.
• Drafts written by an outside agent are now labelled as agent-written instead of as ActVox's own AI.
• The ActVox CLI checks for new versions in the background and updates itself with actvox upgrade.
• The Help Center now explains 3D model blocks in all four languages.

Fixed

• Pages that agents create now appear in your sidebar right away, even in workspaces with many database rows.
• Agents listing your workspace tree or long lists now get every item, including ones edited while they read.
• Filters, sorts and totals on calculated columns now use the real values instead of returning nothing or zero.
• A saved filter on a calculated column no longer hides every row; if it cannot apply, a banner says so.
• Database exports now keep the view's sort order.
• Importing rows that link to thousands of other rows no longer fails.
• Sums and averages no longer show stray digits such as 5086715.400000002.
• An agent can no longer mark a post as published without it actually being posted.
• Archiving a content item keeps its calendar placement, and restoring it puts the item back on the calendar.
• Selected and highlighted text is readable again in dark mode.
• Choosing a USDZ file first no longer leaves an empty 3D block, and 3D file errors now say what went wrong.
• The install command in the Connectors tab now works on Linux servers, and actvox.com/releases opens the changelog.

Fixed

• Home-screen shortcuts, link previews and app connectors can now pick up the current ActVox logo instead of the old one.

New

• A new Memory space keeps the facts your workspace should remember, each with its subject, its source and how sure it is.
• Turn on recall in Settings and the assistant brings the relevant remembered facts into each chat.
• Turn on the session digest and the assistant suggests new facts after a conversation for you to accept or reject.
• Remembered facts appear in knowledge search and in the knowledge graph, linked to what they are about.
• One ActVox CLI sign-in can now cover several workspaces; switch with actvox workspace use or target one command with --workspace.
• Database tables can show a count, sum, average, minimum or maximum under each column.
• Changing a column's type now shows how many rows change and which values would be lost before you confirm.
• Agents and the API can add rollup columns, export a database to Excel, and import or export up to 1,000 rows at once.

Improved

• Database filters add greater than, at least, less than, at most, is any of and is none of.
• Importing a CSV into a database lets you choose Strict or Lenient and gives you a per-row report you can download.
• The page where you approve a connection now names exactly what it creates for each workspace you select.
• The CLI sign-in confirmation is now a clear ActVox page in your language.
• Links to os.actvox.com now show a proper preview image when you share them.

Fixed

• A database cell with an unknown option, a malformed date, email or link is now refused with a clear message instead of being saved or silently dropped.
• Editing a select column's options through an agent no longer wipes them.
• Formula columns now show their values everywhere, including exports and the API.
• Removing a database column now removes its old values too.

New

• Checkout now describes each plan and what it includes before you pay.
• On the Free plan, the sidebar shows how many of your workspace's 25 pages you've used, with an upgrade option when you're close.

Improved

• When you cancel a subscription, the billing portal asks why — one click, and it helps us fix what's missing.

Fixed

• Rewording or clearing text in an HTML block's Edit mode now changes the HTML itself, so Source shows your new text.
• Changing text in Edit mode and then changing it back no longer makes it flicker between the two versions.
• Shared pages, exports and the assistant no longer show older text after you reload a page right after editing it.
• A Team plan that lowered its seat count now receives the right monthly credits at renewal.
• A plan downgraded for the end of the billing period now switches to the lower plan at renewal instead of keeping the higher one.

New

• Home is now your workspace's front door: with no tab open you land on it, with a box to start a chat, your recent work, your skills, your connectors and quick Create buttons.
• Type / in the Home chat box to pick one of your workspace's skills or commands; the skill cards on Home run them in one click.
• Each person now keeps their own database view — layout, grouping, filters, sorting and columns — and can reset to the team default or make theirs the default for everyone.
• Shared pages open the pages inside them, with breadcrumbs back to the start, and show embedded databases instead of a placeholder.
• You can comment on a selection that spans several paragraphs, and the comment box now opens right under your selection.
• HTML blocks have Preview, Edit and Source modes: fix text directly in Edit, and use Format source to make minified code readable.
• The sidebar shows which version you're running, and a banner tells you when a new version is ready so you can refresh.
• An invitation sent to a mistyped address, such as gmailc.om, is cancelled automatically, and you're shown the address you probably meant.

Improved

• The sidebar is easier to read: larger text, stronger contrast, and an arrow that appears in place of a page's icon when it has pages inside.
• Recent shows each page's own icon, keeps its Today, Yesterday and This week sections, and its filters open in a compact menu beside the button.
• Instagram, LinkedIn and X drafts get more time to finish, so fewer come back empty.

Fixed

• Connector logos such as X, GitHub and Notion no longer disappear in dark mode.
• Database headers no longer wrap their buttons onto two lines, “+ New row” is readable in dark mode, and only one menu opens at a time.
• Board columns no longer scroll sideways underneath their cards.
• Shared pages follow the page's width, size HTML blocks to their content, show interactive galleries instead of a blank frame, and no longer show the app's header.
• The text toolbar no longer hides behind the page icon.
• Edit mode now saves text in HTML blocks whose content is drawn by their own scripts.
• Starting a chat from Home no longer loses your first message.
• A workspace's calendar can now only be read by people with access to that workspace.

Fixed

• Connecting ChatGPT, Claude or Grok to several workspaces failed with “Explicit write consent is required” when one of the ticked workspaces already had write access; approval now goes through.
• A permission that only some of the ticked workspaces already hold is shown as a choice — “Previously approved on …” — so you decide whether the other workspaces get it too; the workspaces that had it keep it either way.

Improved

• Reports finish in about half the time — roughly ten minutes instead of seventeen.

Fixed

• A report could list another company's social accounts as yours — for example a logo directory's Instagram — and write a content plan for it; only accounts linked from your own site, or clearly named after you, are used now.

New

• The consent page lets you tick every workspace the assistant may use, grouped by Personal and by organization, with Select all; the first ticked workspace is the primary and each one stays separately pausable and revocable in Settings.
• An assistant connected to several workspaces can name the workspace it wants on any request; requests without one go to the primary, and “capabilities” lists the workspaces the connection can reach.

Improved

• Disconnecting from the assistant’s side removes every workspace of that connection at once, while revoking one workspace in Settings leaves the others connected.

New

• When you connect an assistant you now choose its permissions right away: read is always on, and “Grant full access” adds write, run and share in one tick, so you no longer have to connect twice.
• The consent page lists every workspace you administer, across all your organizations, with the organization named under each one.

Improved

• The consent page uses ActVox’s own dropdown and checkboxes and follows your light or dark theme.
• The Grok setup card explains that Grok registers itself, and only asks for a client ID if Grok’s form does.

Fixed

• Grok’s connector setup failed with “Connection failed”; Grok can now connect.
• Organization admins were offered only their personal workspace when connecting an assistant.
• The floating Support button no longer appears on the consent page or on shared pages and reports.
• AI assistants now show ActVox’s current logo for the connector instead of the retired green mark.

Improved

• Refreshed reports and reports your assistant starts now resume after a platform interruption and can be cancelled, like any other report.
• A report that stalls is closed out after an hour instead of an hour and a half.

Fixed

• Some finished reports came back without their Instagram and LinkedIn content plans; they are now included.

New

• Export a page as PDF or PNG from its header, with paper size, orientation, margins and fitting options and a live preview; Word, Markdown and plain text exports are made in your browser.
• Export a database as Excel or CSV from its header.
• Interactive blocks print as you see them in PDF and PNG exports, instead of a line of fallback text.
• The sidebar's New menu creates a chat, page, database or whiteboard from one place.
• A report that stops making progress tells you so after ten minutes, with Cancel right there, instead of spinning until it is failed.

Improved

• Dark mode is recalibrated: tabs, buttons, chat bubbles, hover states and status colours keep their contrast, and whiteboards follow the theme instead of opening white.
• Light mode text uses a softer reading weight, and warning and info colours meet contrast guidelines.
• The “Reconnecting to your workspace” notice waits out a brief blip before it appears.
• Reports run their audit lookups in parallel and pick up where they left off after a platform retry, so they finish sooner and more reliably.
• The frameworks and libraries behind the app are up to date.

Fixed

• Claude, Claude Code and Grok connectors can be approved again; since v3.2.0 they were refused at the consent step.
• Cancelling a report now stops its run instead of letting it keep working in the background.
• A finished report no longer gets stuck at the last step when saving its results is slow.
• In dark mode, inactive tabs were unreadable, primary buttons inverted and your own chat bubble was white on white.

New

• A Copy button in the header of every page, database, and whiteboard copies it as Markdown, with plain text in its menu.
• Code blocks on a page show a Copy code button on hover that copies the whole block.
• Every database has a New row button in its header, so you no longer scroll to the end of a long table.
• Databases can run at standard or full width, like pages, from the More actions menu.

Improved

• Pages, databases, and whiteboards share one header: the title, quick actions on hover, and Copy and save status on the right.
• Database views switch with tabs centred above the grid, and a table narrower than the view sits centred under its title.
• Like and dislike rest as a single thumb until you hover over them.
• Adding a cover is one button, with Upload image and Choose existing in the same menu.
• Reports finish sooner: timed-out audit sections are retried together, and Instagram, X, and LinkedIn drafts start on the model that completes them.
• The ActVox logo and favicon are refreshed across the app, the website, and sign-in.
• Whiteboards load their fonts from ActVox instead of a third-party server.

Fixed

• Reports generate their Instagram, X, and LinkedIn content plans again; since late July those sections came back empty.
• Apply report works when a report is missing one social draft, instead of hiding the button.
• A report that stalls now fails within about 20 minutes instead of spinning for an hour or more, and can be cancelled in any phase.
• One slow step no longer wipes a report's whole content strategy; only that section is lost.
• The Cross-Platform Strategy section of reports is filled in again.
• Redo on a whiteboard no longer loses your text after a teammate opens the same board.
• A database table no longer covers the left edge of its pane.
• Affiliate referral links take new customers straight to sign-up instead of the homepage.

New

• Stripe Checkout now handles subscriptions and one-time ActVox Credits, with a Customer Portal for plans, payment methods, cancellations, and invoices.
• Affiliate referrals can be attributed directly from Stripe payments without duplicate conversion reporting.

Improved

• Billing clearly shows personal or organization payment scope, credit balances, and tax information before checkout.
• Media Studio previews keep their selected image synchronized across device previews and editing state.

Fixed

• Website reports now include brand social profiles linked from page footers, not only links in the main page content.
• Legacy Clerk billing controls no longer compete with active Stripe subscriptions.

New

• Content now accepts still-image sources up to 100 MiB, converts and resizes them before upload, and keeps stored images within publish-safe limits.

Improved

• Instagram stills are prepared as JPEGs at or below the platform's 8 MiB limit, while video, PDF, and multi-asset rules stay unchanged.
• Website reports preserve social profiles linked from the submitted site, skip platforms without verified brand accounts, and keep genuine verified-account failures visible.

Fixed

• Search-only founder and personal profiles no longer appear as a company's social accounts or make a report Partial by themselves.
• Connector availability now reflects current approval evidence for hosted Claude, Claude Code, and Grok clients.

Fixed

• A report missing its Instagram, LinkedIn, TikTok or X drafts now tries a second AI model before giving up, so one slow provider no longer costs you those drafts.
• A platform counts as recovered only when its drafts are really there, so a report can no longer read Complete over an empty content section.
• A Partial report turns Complete once you accept the regenerated section, and until then the banner names the part that is still missing.
• Copying a page brings its nested whiteboards with it — the boards keep their drawings and their links to the copied pages — and deleting a page or board now clears its collaboration data too.
• Table controls in the page editor are clearer: separate buttons for adding a row and a column, delete actions in each row and column menu, and no empty gutter when the controls are idle.

New

• A report whose SEO, competitor or strategy section came back empty — or whose social or platform-content step failed — now shows Partial instead of Complete, with the finished sections still readable and the run refunded; press Regenerate on the empty section, and Apply report appears once the report is whole.
• Report scores show N/A when the scored analyses returned nothing, instead of a zero and a poor rating that nobody earned.
• A chat request now stays pinned to the page or report it was written on: the Editing chip (for a page) or Context chip (for a report) is saved with the message and comes back after a reload or reconnect.
• Agent bundle 3.1.1: creating a Page through the CLI, SDKs or MCP always makes a new one, and an explicit update must be based on a current read, so an agent can no longer replace a page it has not looked at.
• A new guide on the blog walks through connecting ChatGPT to your workspace with the ActVox MCP connector — five steps from the connector card to the first grounded answer.

Improved

• For external agents: an update through mode=upsert without the page's current ETag now answers 409 artifact_diverged instead of overwriting; the 3.1.1 CLI, SDKs and skill send the ETag for you, so update the bundle if you script Page updates directly.
• The browser tab reads ActVox, not the full product tagline, so a row of workspace tabs is legible again.
• The agent refuses an edit whose target page was closed, retired, or belongs to another workspace before it spends anything, and says so.

Fixed

• Editing an inline comment keeps the editor on that comment's row, with its author and time still visible, instead of jumping the composer to the end of the list.
• A report that recovered its social sections no longer keeps a failed health warning after it is complete.
• Older reports that were marked complete without their required sections, or with a failed social step, now display Partial and no longer offer Apply report until they are repaired.

Fixed

• Instagram feed carousel images now use one consistent 4:5 portrait format, so stored assets and previews no longer change shape.

New

• Every Help section page has Copy page, which copies just that section as Markdown, and View as Markdown, which opens the section as plain text for an agent or a document.
• Help → Automation and connections now says, per agent, whether the MCP connection and the skill are available for ChatGPT, Claude, Claude Code, Codex, Hermes, Grok and OpenClaw — in a table you can scan on a phone or in Arabic.
• The MCP and skill cards and Help carry a read-only Knowledge Base smoke test: a first prompt that names the workspace and lists three sources, and refuses to store anything about you until you confirm the exact fields.
• The agent-ready section on the home page links straight to the connector documentation, next to the integration setup and the npm CLI.
• The home page's Workspace OS section now explains the content workspace — Your ideas. Your pages. Your voice. — with four benefits (flexible pages, AI writing that feels human, your brand DNA built in, one workspace for your team) and a Start creating with ActVox button that opens onboarding.
• When a report's strategy failed to generate, the Strategy page says so and offers Regenerate strategy; once the new strategy lands, Apply report appears as usual.
• A report whose Blog section came back empty now keeps its header and Regenerate button, and says whether the section was never generated or failed.

Improved

• Agent bundle 3.0.0: the research-actvox skill name, the ResearchActVox client aliases and the RESEARCH_ACTVOX_API_URL, RESEARCH_ACTVOX_API_KEY, RESEARCH_ACTVOX_STATE_DIR, ACTVOX_BASE_URL and ACTVOX_TOKEN variables no longer work. Reinstall the bundle from Skills and Connectors, call the skill actvox-workspace, and set ACTVOX_API_URL, ACTVOX_API_KEY and ACTVOX_STATE_DIR.
• Saved prompts and scripts that mention research-actvox need updating; the installed bundle now contains only actvox-workspace, actvox-document-studio, actvox-html-artifact and actvox-ws-dna.
• Strategy and content-concept generation get twice as long to finish, so a longer report no longer times out at the step that feeds the Blog.
• Help's ChatGPT, Claude and Grok guidance uses the same documented, not available and blocked wording as the connector cards, and keeps the profile, instructions and Knowledge Base stores separate.
• The home page no longer carries the illustrative example report or the See an example report button; the hero has a single Start my free audit action.

Fixed

• The workspace switcher at the top of the app shows the active workspace's name and logo — not the organisation's — and keeps up when you switch between workspaces of the same organisation.
• After a strategy timed out, the Blog section had no Regenerate button to recover with; it does now.
• The I don't have a website - chat with the assistant button on the home page pointed at an old workspace address and lost where you came from; it now opens onboarding directly.

New

• Add a 3D model to a page: type / and choose 3D model, upload a GLB or glTF (plus a USDZ for AR on iPhone and iPad), then orbit, zoom, auto-rotate, go fullscreen, capture a poster from any angle or download the file.
• Source on an HTML block opens a code pane beside a live preview of your draft — a draggable divider, line numbers, Tab to indent, ⌘S to save — and interactive artifacts get a Details table for title, height, data and fallback text.
• HTML blocks and interactive artifacts gain Fullscreen, Open in new tab (a page of its own with just the artifact) and a Hide header option.
• Connect ChatGPT, Claude (claude.ai, Desktop, mobile and Cowork) or Grok to a workspace from Skills and Connectors — each tile carries the whole journey: add it in the assistant, allow one workspace on the consent page, use it in a chat, remove it.
• The copyable first prompt now onboards the agent: it checks what it can do, reads what already exists, interviews you, and — only after your yes — stores your business portrait in the profile, its behaviour rules in the instructions and your sources in the knowledge base.
• Help → Integrations gains “Connect ChatGPT, Claude or Grok (MCP)” with the same journey in prose, team-plan notes and what “client not approved” means.
• The actvox-workspace agent skill 2.1.0 adds operating rules (read before write, propose first, ask before sharing, publishing or spending credits), a reference of the blocks an agent may write, a context command and clearer errors.
• The Claude Code, Codex and Hermes tiles show how to update the skill and how to set it up on a remote host over SSH with just the workspace key — no browser needed.

Improved

• A linked database view inside a page shows the live grid instead of a placeholder card; switching Table, Board, Calendar or Gallery on the embed changes only that block, and /db now offers Database - Link existing.
• A database created inline with /db nests under its page in the sidebar, like a full-page database.
• The database table view sits in one frame with hairline rows, readable text, mono column headers and one-line cells — long notes and URLs no longer make rows ragged.
• Column and row grips stay faint until you hover, and always show on touch.
• HTML blocks draw one frame and one toolbar — Preview and Source as chips, the title once — while the share page keeps its card.
• The command-line tile is now called Claude Code; Claude is the chat app. Claude Code and Grok connections are enabled under a time-boxed review window.

Fixed

• Collaborative pages no longer lose blocks after a reload or an editor remount, a new page no longer deletes the first words you type, and edits made after a code reload are no longer stranded — where they were, they are kept as a version you can restore.
• Restoring a version on a collaborative page no longer duplicates every block.
• Name dialogs no longer reset what you typed while the page autosaved — a child page named in the prompt is no longer created as Untitled.
• Connecting Claude no longer ends on “This connection is unavailable or expired”; when a client really is not accepted, the consent page says so and names it, without a Retry that could never work.
• In the HTML source editor, clicking the toolbar can no longer delete the block, focus lands in the code pane on open and returns to Source on close, Ctrl/⌘+M toggles whether Tab moves focus, and save errors show under the code pane.
• Replacing a 3D model clears the old poster, and Download saves the file under its real name instead of navigating away.
• An agent writing raw HTML blocks through the API could bypass the interactive-HTML limits; those blocks are now refused, and the skill's reference lists exactly which blocks an agent may write.
• The skill's verify command uses python3 (the previous command failed on machines without a mise Python), reinstalling after running the skill in place no longer fails on interpreter caches, and table-of-contents blocks now get an id so agents can address them.

New

• Every Studio lane — Edit, Image, Carousel and Video — previews the post on the same phone, with the platform's own chrome around your text and media.
• Video previews follow the platform: a full-screen Reel or TikTok stage for Instagram and TikTok, and the video inside the post for X and LinkedIn.
• The preview folds into a drawer — hide it to give the editor the whole pane, show it again with one press; the choice is remembered on your device.
• The caption under the phone is counted against the platform's own limit (280 for X, 2,200 for Instagram and TikTok, 3,000 for LinkedIn) and saved with Save caption.
• Edit a video concept's script — Hook, Scene description, Visual style, Text overlays, Call to action and Audio — before you generate; edits save as you type, the concept is marked Edited, and the video renders from your version.
• Concept chips lead with their pattern — Curiosity, Problem→Solution, Result Reveal, Hot Take — with the hook as a one-line tail.
• A finished video says Video ready, shows Saved to this post when the draft already holds it, and otherwise offers Use this video.
• The Studio header shows the draft's title, platform and status, and whether all changes are saved; the lane tabs switch with the arrow keys.
• A slim status line replaces the always-open activity rail: one line for this draft's current or just-finished run, with View activity to unfold the cards.

Improved

• The Image, Carousel and Video lanes share one layout: the controls on one row, the cost under them ("4 variants = 300 credits", "N slides × 75 credits", "5 s = N credits"), the prompt, then one primary button.
• In the Carousel lane, Redo slide is the first action on every slide, and the previous slide stays visible while its replacement renders.
• The Video lane's controls read "5 s", "9:16" and "720p", with the detail in the list, and show which brand style is applied.
• Your agent's ActVox skill is now called actvox-workspace (bundle 2.0.0). The old research-actvox name keeps working as an alias for one release and prints a reminder; the settings are ACTVOX_API_URL and ACTVOX_API_KEY.
• The connector setup, the copyable agent prompt and the verify commands all use the new skill name.

Fixed

• LinkedIn drafts previewed as a desktop card while every other platform was on a phone; LinkedIn now previews on the same device and its video sits inside the post.
• Every platform's video showed the same generic black frame; the Reel, TikTok, X and LinkedIn previews now each look like their platform.
• The phone preview was a bezel sized to the media — a black square around a grey box — instead of a device showing the post.
• The Edit lane still showed a different, bordered preview after the other lanes moved to the drawer.
• Two buttons could look primary in the Video lane at once; only one action is primary in each state.
• Use this video could do nothing when pressed on a video the tab had already seen; it now saves every time and confirms with a toast.
• Script edits did not show as unsaved changes while they were still being written, and an edit made just before leaving the page could trigger a Leave site? prompt.
• A saved video prompt was cut to the limit of the model selected while typing instead of the model that renders it.
• Concept chips stacked one per line, the script fields opened with warning counters on normal output, and the model dropdown sat above the brand cell.
• The Video lane's control row wrapped its brand cell onto a second line.
• Connecting Google Search Console is bound to the site that started the sign-in, so an authorization begun on one environment cannot be completed on another.
• Reinstalling the agent skill on a machine where it had been run once was refused as a local edit; reinstalls work again.
• Website scrapers and the agent's product context pointed site owners and agents at the retired research.actvox.com address; they now name actvox.com.

New

• Install the ActVox CLI with `npm install -g @actvox/cli` — it is published on npm, works on macOS, Linux and Windows with Node.js 20 or newer, and `actvox --version` confirms the install before you sign in.
• The Codex, Claude Code and Hermes connector setup starts with an Install step: pick npm, pnpm, yarn or bun and the command and its Copy button follow your choice.
• The homepage's agent-ready section shows the real install command with the same package-manager tabs, a copy button, and a link to the package on npm.

Improved

• The ActVox CLI connector's official link goes to the npm package page instead of a page that did not exist.
• The public agent context (llms.txt) describes the CLI, how it signs in, and that profiles start read-only.

Fixed

• Privacy Policy, Terms of Service and Cookie Policy no longer advertise Arabic, Russian and Ukrainian versions that do not exist, so search engines see one consistent English page.
• Files & Media action buttons — detach, delete, upload and remove — render at their intended size and colour instead of as full-width primary buttons.
• In Arabic, dragging the Studio splitter moves the preview in the direction you drag, and the handle has a translated label.
• A running or failed operation no longer covers the video model picker, the brand summary or the carousel heading with a banner you could not click through.
• After one failed generation, the other Regenerate and Remove buttons no longer announce themselves as failed to screen readers.
• The prompt box no longer clips your text after you resize the Studio split.
• A video concept that could not be saved now says so instead of reporting that the video could not start.

New

• Image, carousel, video and text generations keep running after you close the tab or reload, and reopen showing their progress or their result.
• A content draft lives in one Studio tab with Edit, Image, Carousel and Video lanes, instead of separate tabs for editing and generating.
• Pick a model preset — Auto, Fast or Quality — or a single provider, and see the estimated cost next to it before you generate.
• Upcoming models are listed greyed out with the reason, so you can see what is coming without choosing something that would be refused.
• Every paid generation records which model you asked for and which one actually ran.
• Generating a new carousel over an existing set asks first, keeps the previous set restorable for 7 days, and shows a version history you can restore from.
• A generated text draft arrives as a proposal you Accept, Compare or Reject — your draft is not touched until you decide.
• The Studio activity rail lists every running and finished generation in the workspace, with an Open action that takes you to its tab.
• The finished-video notification opens the Studio tab instead of a download link that expires.
• The Brand logo control in the Studio shows the logo that will be applied, with an Edit link into Brand DNA.
• Manage cookies and export your workspace data from Settings → Privacy and data.
• The ActVox CLI signs you in through your browser and keeps its credentials in your system keychain.
• A Startup Fame badge joins the badges in the site footer.

Improved

• Privacy Policy, Terms of Service and Cookie Policy rewritten from what the product actually does, with a complete cookie inventory.
• Terms and Privacy links now appear on sign-in and sign-up.
• Video generation moved to the provider's current model ahead of the old one's retirement, and images moved to the newer Grok Imagine model.
• The Video lane follows the same layout as Image and Carousel: model first, then format controls, then concepts.
• Uploads and attachments in the content editor show their progress in place instead of only in a toast.
• Cookieless visit analytics on actvox.com.

Fixed

• Reloading during an image render no longer fails it and refunds the credit — the render finishes and is there when you return.
• Long image and carousel renders are no longer cut off and refunded partway through.
• A video whose saved copy failed to land is no longer left as a broken link — saving is retried, and the provider's copy is kept if it cannot be.
• Regenerating images no longer blanks the ones you already have while the new set renders.
• Generating text no longer overwrites your draft, and closing the tab no longer loses the run.
• The first rating after reopening a Studio session is no longer lost.
• Every Studio media lane shows the editor and preview side by side on wide screens instead of stacking them.
• The Studio tab no longer offers Generate Image and Generate Carousel twice.
• A draft's Edit tab could vanish on reload; the single Studio tab now keeps every lane.
• Reopening a finished video no longer overwrites a video you uploaded afterwards.
• Detaching a carousel slide keeps the slide count right, and detaching the last one lets you generate a fresh set.
• Choosing an unknown video model is refused instead of silently switching to Auto and billing you.
• Deleting an organization now removes everything that belonged to it, including connected-account tokens.
• The cookie-consent cookie is now sent only over secure connections.
• Platform names stay visible next to their priority in Strategy → Platform Priorities.
• Generated document previews render Markdown safely, Copy as .md exports the full latest source, and History stays available while editing.
• The favicon and sign-in logo show the current mark, and the example-report button on the homepage is readable over the hero image.
• Interactive HTML artifacts no longer come up blank in the page editor or on shared pages.

New

• Every save of a generated document is kept as a version, with who saved it and when.
• Preview any earlier version side by side with the current one before you commit to it.
• Restore an earlier version in one step — it is added as a new version, so nothing in the history is lost.

Improved

• A content draft and its Media Studio now share one tab instead of opening two for the same piece.
• Connected AI clients now see accurate warnings about which actions overwrite content or spend credits.

Fixed

• The Content library remembers your filters, sort order, search and layout when you switch tabs and come back.
• It also returns you to where you had scrolled to, and to the draft you had open.
• Coming back to a tab shows the content you already had instead of a full-page loading skeleton.
• Two browser windows on the same workspace no longer move each other's selected tab.
• Switching a Media Studio between image and video, or moving between Skills sections, now survives a reload.
• A tab can no longer show one item's heading above another item's content.
• Deleting a page comment now asks you to confirm first, and keeps the comment with a clear message if it fails.

New

• Connections in Settings lists every AI client connected to your workspace, when it last called and what it verifiably changed.
• Pausing or revoking a connection is now recorded: the client's next call is denied and shown with the reason.
• Copy evidence exports one record of a connection — its verified identity, consent timeline, every call and denial, and its last verified effect.
• Every AI client now has its own card with copyable setup steps and a badge saying whether its connection is verified or not available yet.
• A tab holding unsaved edits or a running generation is no longer closed to make room for a new one.
• When every open tab is protected, a dialog lists each one and its state so you choose what to close.

Improved

• The permission screen names the client actually asking for access and starts with write switched off, so read-only is the default.
• The permission screen now reads in English, Russian, Ukrainian and Arabic.
• A generation you leave running in the background keeps counting as running, and its tab says so rather than looking idle.

Fixed

• Signing in partway through connecting a client no longer starts a second request — the original one resumes where you left it.
• Reloading the permission screen keeps you in the same request instead of sending you back to the client to start over.
• The authorization link no longer lingers in your browser history with its sensitive details.
• Reconnecting a client whose connection you paused now tells you it is paused, instead of quietly turning it back on.

New

• Video concepts are saved to your workspace instead of being lost when you leave the tab.

Improved

• A new landing page leads with the free audit and the three steps from website to plan.

Fixed

• Generated images now stay visible after switching tabs, instead of vanishing from the panel.
• Asking AI to write a LinkedIn, X, Instagram or TikTok post now returns a post for that platform rather than a long article.
• Video generation follows the platform of the draft you are working on instead of always assuming TikTok.
• When a platform draft cannot be produced, you are told the actual reason and are no longer charged for an empty draft.

Fixed

• Generating an image no longer fails at the last step after the image was already created.

New

• The home page now leads with a free audit and shows the three steps from website to marketing plan.
• A worked example shows how a single finding turns into evidence and a recommended action.
• New guide: what an AI visibility audit is and what it covers.

Improved

• Pricing now states each plan in ActVox Credits, with estimates for audits, text, images and video.
• Image and video generation now runs through one service instead of each screen calling providers on its own.
• If a generation is interrupted mid-request, it is never silently sent a second time or charged twice.
• A video that was already submitted is recognised on retry instead of starting a duplicate job.

New

• The site footer now links to our verified Instagram and YouTube channels.

Fixed

• Content Calendar, Content Plans and Social Posts now show the posts and plans your report produced, instead of empty cards.
• Social Presence shows guidance to rerun a timed-out audit instead of asking you to connect social accounts.
• The Competitors section describes which sections fell short instead of repeating the same agent name.
• Image carousels complete every slide, even on ten-slide sets, instead of timing out partway.
• Declining a suggested page edit no longer ends the turn in silence — you get the assistant's clarification.
• Setup no longer asks the profile and setup questions together, or repeats a question it already asked.
• Setup briefs now complete on realistic company profiles instead of saying a draft couldn't be prepared.
• Share and scheduling requests in your first message now follow their own flows instead of rewriting the page.
• The text-formatting menu closes when you switch or close a tab, click away, or press Escape.
• Completing setup shows the suggested next step even when you built something other than content first.
• Long assistant turns keep running after a dropped connection, and you can rejoin the finished result.
• The assistant's quick confirmations now arrive in your language — English, Russian, Ukrainian or Arabic.

New

• Connect ChatGPT, Claude Code or Grok to your workspace over MCP — each connection consents in your browser with the permissions you approve, and you can pause or revoke it in Settings at any time.
• Duplicate a page someone shared with you into your own workspace as an independent copy — the source page and its workspace stay untouched.
• The first-run setup interview starts on its own while your company report runs, and your answers continue the interview in progress.
• Workspace setup now speaks English, Russian, Ukrainian or Arabic — approved titles, schemas and defaults compile in the language you chose.
• Setup reuses what it already knows from your project and report, asks fewer repeated questions, and fills a first brief with grounded ideas, source notes and a measurement plan.
• Brand DNA, DESIGN.md and ACTVOX.md are available as standing resources through the API, CLI, SDKs and MCP.

Improved

• The seeded first-run chat tab is now named ws-DNA in every language.

Fixed

• The notification bell now keeps the badge, the unread list and “mark all as read” on the same consistent set.
• Answering a setup question mid-interview no longer ends the turn without a reply — the remaining work continues.
• The next setup question stays visible after your profile is saved, and a finished setup keeps its verified answer and first-stop link on screen.
• Sharing links scoped to your organization from a personal workspace now return clear guidance instead of a broken flow.
• Loading indicators in chat stay scoped to the active turn, and finished messages settle correctly without a reload.
• Drafted setup content is checked against the facts you supplied, and unsupported claims are rejected.

New

• Connected tools can turn saved content or generated artifacts into editable pages, append them to existing pages, or refresh a source block.
• Connected tools can create nested pages and databases together, with retries that avoid duplicate structures.

Improved

• We are gradually enabling workspace and organization switching that keeps the surrounding screen in place.
• In the new shell, separate windows keep their own workspace URLs, and moving chat between panels retains the same conversation runtime.
• The assistant and report pipeline use an updated model; your existing chats and reports remain available.

Fixed

• Opening the same newly created page in two panes no longer risks erasing its initial text.
• An unavailable tab stays inside its pane, preserving the sibling editor and unsent chat draft in the new shell.
• In the new shell, chat streams stop cleanly when leaving an organization, while completed answers remain available in their original workspace.
• Approval buttons wait for the current chat turn to finish before accepting another decision.
• Updated the application and sign-in dependencies, including the image-processing security update.

Fixed

• Accepting an assistant suggestion on a page no longer fails with a server error; “Accept all” and per-change acceptance both apply.
• Suggestions that add or change numbered lists, links or aligned text no longer fail on acceptance with a “does not match the snapshot” error — the page keeps the change instead of rolling it back.
• A suggestion that replaces a short page with a long one now shows whole lines removed and whole blocks added — headings, bullets and table rows on their own lines, each with its own Accept and Reject — instead of strike-through fragments mid-sentence.
• Blank lines the editor adds on its own are no longer offered for review.
• “Add 3 more tasks to the existing To Do List” adds the tasks to that database, found by its name or because it is the database you have open, instead of declining.
• The assistant’s own offers (“tell me what to make and I’ll create it”) are no longer mistaken for claims of a change it did not make.
• “Refactor this page text” and “make better formatting” now draft a page suggestion instead of answering about the page.

New

• Ask the assistant to add information to a database that already exists — “add this information to Task tracker” — and it drafts new rows for you to review instead of declining.
• Ask it to fill, finish or update the page you have open, in English or Ukrainian, and it drafts the change as a reviewable suggestion on that page.
• While the assistant works, the typing indicator names the step in progress — “Drafting variants”, for example — instead of a silent pause.
• Edit an API key’s scopes after it is created: every active key under Settings → API keys has an “Edit scopes” action, and the key itself stays the same.
• “Select all” and “Clear” in the API key dialog fill or empty the scope list in one click; “Select all” leaves out publishing, which needs a key of its own.

Improved

• Content drafts arrive faster: the three variants are written in parallel instead of one after another.
• Follow-up questions in a long conversation answer faster; the assistant reads the recent turns rather than the whole thread every time.
• All inline comments on a page are listed together while their markers stay in place.

Fixed

• Asking for a to-do list in a database no longer produces a blog draft.
• The assistant no longer says it created or updated something when nothing was written; it tells you exactly how to ask for the change instead.
• Creating a page together with its text no longer detours through a research step that asked you to re-send the request.
• Sharing a page with your organization after the assistant asked which page and audience now stays in the sharing flow.
• The API key dialog explains why publishing cannot be combined with content-writing scopes before you press Generate.
• Affiliates whose commission rate was set under an older program are now paid the current rate, and pending and paid earnings match the affiliate program’s own records.

New

• Manage Account now opens your whole account: profile and email addresses, connected sign-in accounts, passkeys, every active session and account deletion, with a Credits page beside them.
• Add a passkey and sign in with your face, fingerprint or device PIN; the devices list shows every session signed in to your account and lets you end the ones you do not recognise.
• Manage organization now covers the name and logo, members, invitations and roles, leaving or deleting the organization, shared credits and seats, and billing — admins edit, members read.
• The Billing page names who is billed and on what, and a team member can switch to their own personal billing without leaving the team workspace.
• Team is priced per seat: pick the number of seats on the Team card and see the monthly total before checkout.
• Payment methods, statements and invoices open from the Billing page in one click.
• Education access is an application, not a purchase: the pricing page and the Billing page offer “Apply for education access” instead of a Student checkout.
• Every balance, pack and plan now says ActVox Credits; the word “tokens” is gone from customer screens in all four languages.
• Account, organization and billing screens follow your language and the dark theme.
• Affiliates see real pending and paid earnings in Refer & Earn, drawn from what the affiliate program will actually pay.
• A refunded credit pack is clawed back automatically once refunds are wired to Stripe, and a refunded sale voids its affiliate commission.

Improved

• Cancelling a subscription keeps your plan until the end of the period you paid for; the downgrade happens only when that period ends.
• The Billing page is denser: one row per fact, one primary action, compact metric tiles and plan cards that fill the width.
• The Billing page shows whether ActVox’s record of your subscription matches the payment provider’s, and a daily check repairs any drift.
• Every billing and affiliate event is written down before it is processed, retried on failure and escalated to a person when it cannot be applied — nothing is silently dropped.
• Blog covers on the marketing site are now high-resolution.
• The Canva connector shows Canva’s official icon.
• The Auto-adapt setting uses the same checkbox as every other setting.

Fixed

• A payment that arrived before your account finished setting up is no longer lost; the credits land as soon as the account exists.
• Buying a personal plan from inside a team workspace no longer changes the team’s plan.
• A second partial refund on a credit pack is no longer ignored.
• Adding seats to Team mid-month grants only the extra seats’ credits instead of a second full month.
• An affiliate commission recorded at the wrong amount is flagged for review instead of being counted as delivered.
• Referral links from affiliates approved in the affiliate portal now bind on the first visit instead of being forgotten.
• Leaving or deleting an organization from its settings no longer leaves the workspace pointing at the organization you just left.
• A per-seat plan whose seat count changed on the provider’s side is now reflected in the seats ActVox shows.

New

• Pages, databases, decks, images and share pages the assistant makes now use your own colours, type and style.
• A workspace with no brand captured yet gets a clean neutral look instead of ActVox’s own.
• You decide when your logo lands on generated images and decks — on request, always, or never.
• Media Studio can stamp your logo on a render on the spot, and tells you what it applied.
• Three brand files let any outside AI tool build on your brand, behind a read-only key you issue.
• Rate any page, deck or image the assistant made, and say what was off when it misses.
• A weekly summary of those complaints now sits in Settings under Engineering.
• Pages the assistant creates arrive with an icon, a cover and a layout suited to the kind of document.
• Databases it creates arrive with an icon, tidier column names and statuses in a sensible order.
• Interactive pages the assistant builds now start on their own, with no button to press first.
• Video prompts now carry your brand’s look and mood in words the video model understands.
• Our FAQ page now answers 44 questions about the product instead of six.
• Three new illustrated articles join the blog, and the GEO versus SEO piece has been rewritten.

Improved

• The logo square in Brand DNA is now the upload target itself — click it to add or replace your mark.
• Language and appearance now share one row in Settings instead of two separate cards.
• Assistant writing follows a tighter house style: no long dashes, no curly quotes.
• When the assistant does not know something it says so plainly, instead of leaving a placeholder.
• It no longer pads a tracker with invented rows, and no longer calls a table-backed section empty.
• Deck PDF exports now follow the deck’s own brand rather than a fixed palette.
• Shared pages now show whose brand they belong to, and their link previews match.
• The homepage footer now carries our Featured on Fazier badge.

Fixed

• The assistant now replies in the language you wrote in, not the one your workspace is set to.
• Your knowledge graph no longer loses or double-counts connections when several sources are indexed at once.
• A message you send while the assistant is waiting on your approval is now answered, not swallowed.
• A very long video prompt is now trimmed instead of rejected, and failures explain themselves.
• When a video provider runs out of credit we say so rather than telling you to try again later.
• If your sign-in was ended elsewhere, the app now asks you to sign in again instead of showing an error.
• A Telegram reply that failed to generate is no longer saved into the conversation as an answer.
• Multi-select tags in database views now respond to the keyboard, and cards no longer open when you tap a tag.

New

• A new Files & Media tab collects every file in your workspace — uploads, generated images, covers, icons and attachments — into one searchable library, and you can upload new files straight into it.
• Filter the library by type or origin, then rename, tag, archive, download or delete any file — deleting warns you where it is still used.
• You can now reuse an existing file anywhere — covers, media blocks, icons, attachments and post composers offer picking from Files & Media instead of a re-upload.
• A guided formula editor helps you build database formulas — pick properties and functions, start from a template, and preview the result on a real row.
• Number columns can display as currency, percent or grouped digits, and date columns as short, long, ISO or relative — everywhere the value appears, including spreadsheet export.
• Spotify artist links are now recognized as an input source when you start an analysis.

Improved

• Cloning a page or database now shows a “Cloning…” state and reports errors, instead of working silently.
• The homepage footer now carries our Featured on Bowora badge.

Fixed

• Renaming a database column no longer breaks the formulas that use it — they update automatically, and removing a referenced column warns you first.
• Deleting a file from one place no longer destroys it everywhere else it is used — covers, uploads and attachments stay safe.
• The calendar's Generate Image, Carousel and Video buttons work again instead of doing nothing.
• Multi-select tags in database views now keep your selection order, show complete labels, and reveal hidden values through a +N popover.
• Typing “12,5” into a number card no longer saves an invalid value — comma decimals, percents and currency codes are understood.
• Database rows with a date and time now appear on the calendar view instead of vanishing from the month grid.
• Images that only ever served as covers or icons now show thumbnails in the library and can be picked again as images.

New

• Databases now carry an icon and a cover like pages — emoji, ActVox glyphs or uploaded images, with framing, shown in the sidebar, tabs, share pages and link previews.
• Icons you upload are kept in a workspace library, so changing an icon never loses the file.
• Integrations get typed change events — every feed entry now carries a stable event id and a precise event type, so reorders and page-body edits are distinguishable.
• The CLI, SDKs and MCP server catch up with the API: create with placement, real tree moves, database row content, and 15 new MCP tools.

Improved

• You can reset a Telegram bot's owner from Settings, so handing a bot over no longer requires reconnecting it.
• Page hover actions are tidier — Add icon and Add cover appear only when something is missing; click an existing icon to change it.

Fixed

• When the assistant says it saved content drafts, they are actually saved — including every variant when you ask for the alternates.
• Pasting links with “add to my knowledge base” now ingests each link and reports what was added, already there, or failed.
• Editing select and status options no longer loses your edit or creates duplicates when options are changed quickly or by several people at once.
• Telegram group allowlists now actually apply, and saving Telegram settings no longer clears your other preferences.
• Two bots in different workspaces sharing a Telegram chat no longer share one conversation.
• The command-line tool no longer treats an omitted move destination as a move to the root.

Fixed

• A freshly created blank page no longer refuses its first autosaves — previously the first half minute of typing produced repeated save errors and a failed-save message.

New

• Your Telegram bot is now private — only you and the people you add under Settings → Telegram → Allowed users can chat with it.
• Anyone else who messages the bot gets short instructions for requesting access instead of an answer.

Fixed

• The Telegram bot actually answers again, instead of replying with an error to every message.
• Bot replies with unusual formatting now arrive as plain text instead of disappearing.
• Switching workspace or organization no longer briefly shows errors on your content views.
• When the video provider is temporarily unavailable, video generation now says so and confirms your credit was not charged, instead of failing generically.

New

• Workspace OS comes in light and dark, chosen in Settings and remembered between visits.
• The dark theme covers the whole shell — sidebar, tabs, pages, boards and the phone navigator.
• The landing page now covers personal brand alongside founder, in-house marketer, agency operator and creator.

Fixed

• A blog article's cover image appears only in the article header, rather than a second time at the beginning of the text.

Fixed

• A shared database now shows every row to whoever opens the link, however much has been deleted from it over time.
• Copying a database copies all of its rows, and the confirmation reports the real number instead of zero.
• Copying a page copies every row of the databases nested inside it too.
• Task reminders and My Tasks keep working on a database that has accumulated a lot of deleted rows.
• Exporting a database to CSV no longer includes rows you deleted, so the file matches what the grid shows you.
• Asking the assistant about a database now gives it every row, rather than a fraction of the board.
• A brief connection problem now says it is reconnecting and recovers on its own, instead of telling you your session expired and asking you to sign in again when it had not.

New

• The use cases library now covers 32 marketing workflows, grouped by the kind of business you run — creator, restaurant, hotel, small business or SaaS.

Improved

• Search now sits in the workspace header next to notifications, so it is one click away from wherever you are — the keyboard shortcut and the command palette are unchanged.
• Auto image generation now prefers GPT Image 2 where it is available, while the four-variant set still mixes in other models.
• The use cases page lays each outcome out in a full-width row, so it reads without squinting.

Fixed

• Connecting LinkedIn or Instagram now confirms your signed-in session before it saves anything, and only someone with access to that workspace can change or disconnect it.
• The pages that finish a LinkedIn, Instagram or Search Console connection now handle whatever text the provider sends back safely.
• Whether a workspace has LinkedIn or Instagram connected is no longer readable by people outside that workspace.

Improved

• The landing page leads with the workspace preview, at a size you can actually read.

Fixed

• Opening a second database row no longer warns about a changed revision and quietly stops saving it.
• Video length offers only the durations your chosen model supports, and a leftover unsupported choice resets itself.

New

• Clone a database and get a real copy — every row, everything written inside each row, its attachments and cover, and the way you had the view set up.
• Clone a page and everything under it: the sub-pages, and the databases nested inside them, all in one move.
• Links between the pages you copied point at the copies, while links to anything outside still point where they always did.
• Add a row from Gallery view, including when the gallery is empty — you no longer have to switch to Table or Board to create one.
• Open a row card as a full workspace tab when you want the whole page instead of the card.
• Choose which field makes your Board's columns from the Board itself, instead of hunting for it in view settings.
• See your active filters as chips above the table, and drop any one of them with a click.
• External tools can now build the same nested structure you can: pages and databases inside a page, moved between sections, in the right order.

Improved

• Opening a tab when you already have seven now closes the oldest one instead of refusing. Nothing is lost — everything reopens from the sidebar.
• Cards in Board and Gallery can be opened from the keyboard, not just the mouse.
• Refer & Earn moved to the workspace sidebar, where you can reach it without opening settings.
• Every article on the blog now has its own cover image, carried through the index and related posts.

Fixed

• "No grouping" on a Board now actually turns grouping off, instead of snapping straight back to the first status field.
• Renaming a field no longer discards an option someone else added while you had the menu open.
• Deleting a very large page or database now tells you it is too big, instead of quietly deleting part of it and leaving the rest.
• Copying a database no longer loses your column alignment.
• Media shared between a copy and its original survives deleting either one.

Fixed

• The generated image brief now arrives in a box you can edit, instead of being fixed text you could only accept.
• Your own wording survives switching between the automatic brief and a custom one, so a draft is no longer lost.
• What you typed is what gets sent — the prompt is no longer quietly recomputed on the way out.

New

• Open your workspace on a phone and the navigator slides over the document, instead of shrinking it into a 75-pixel rail.
• A tab switcher made for a phone: tap the pill at the bottom to see every open tab, switch, close one, or unwind a split.
• Back closes the drawer or the tab switcher, the way it does everywhere else, instead of dropping you out of the workspace.
• The bottom bar names the page you are in and tells you whether your edits are saved.
• A shared database opens as cards on a phone, so each row reads on its own instead of scrolling sideways.
• A shared page or database now shows when it was last updated.
• Follow a share link to something in your own workspace and you land in the real, editable object.
• Give a page any icon: search the full set, pick the ActVox mark, or upload your own image.

Improved

• Shared page and report headers are shorter and line up with the document you came to read.
• A phone opens a database as a gallery, because the table could never show more than its first column at that width.
• The chat button now carries the ActVox mark instead of a generic robot icon.
• Search, chat and workspace settings moved into the tab switcher on a phone, so the top bar is just your workspace and your account.

Fixed

• The database toolbar — view settings, export, import and share — was cut off on a phone and could not be scrolled into view.
• Part of the top bar was cut off on a phone, including the button that opens chat.
• A tab's close button was invisible on touch but still closed the tab when you tapped where it would be.
• Tab menus could get stuck open on touch, with no way to dismiss them.
• The icon and menu popovers now close when you tap elsewhere or press Escape.
• A shared database's "last updated" time now reflects edits to its rows, not just changes to its columns.
• Following a share link while switched to your personal account no longer drops you into an unrelated workspace.

Improved

• Connected social accounts no longer add 50% to the price of a report or a refresh — a connection only lets you publish, it does not give the audit private analytics, so it no longer costs extra.
• The social audit says plainly that its metrics come from public profile data; it never labels them "verified" any more, and the connect prompt now describes what connecting actually does.

Fixed

• Connect LinkedIn on os.actvox.com now completes: the sign-in window used to hand your approval to the wrong environment and the card would spin for two minutes. A failure now shows its reason instead of a spinner.
• When the agent pauses for your approval, the chat no longer shows "The run ended without a matching authoritative assistant checkpoint" — the approval card is the answer and it is saved with the conversation.
• Approve works the first time: the card no longer asks you to refresh and retry, and an approval that had been re-requested after a failed run can be approved again.
• Approving an action no longer ends in a "409 Conflict" error with the reply arriving minutes later; the agent's confirmation shows up right after you approve.
• The agent no longer names a page or report after a fragment of your sentence ("findings as a", "of sizes") when you asked it to create something without giving it a title.
• Telegram bot tokens are now stored encrypted; two settings fields that could expose a bot token to the browser are removed.

New

• Your agents can read and update the workspace's business profile and marketing strategy through the API, with the same conflict check every other edit gets.
• Reports are open to agents: list them, read a summary or one section at a time, override a section's text — and, with the new execute power, re-run the audit for a finished report.
• Content drafts can be created, edited, scheduled and archived by an agent, and the calendar placement travels with the draft.
• The knowledge base is open to agents: add a URL or a text source, re-index it, read its indexed text, search it (free text or the smarter hybrid search) and explore the knowledge graph.
• Agents see the workspace the way you do: a file tree with folders they can create, rename and move things into, a search across every kind of object, and a change feed they can replay from a cursor — deletions included.
• Three new agent powers you grant separately: execute (run allow-listed actions and report refreshes), share (mint and revoke share links) and publish (schedule a native Instagram or LinkedIn post). A key can never hold publish together with a writing scope, so a human always approves what an agent publishes.
• Agents can batch up to 25 requests in one call, each answered on its own.
• A Python SDK (pip install actvox-sdk) and an MCP server (actvox-mcp) plug Claude, Cursor or any MCP client straight into a workspace with one key.
• Knowledge sources and reports can now be archived and restored by an agent instead of deleted — archived items leave every list and search but stay readable by id.

Improved

• Agent traffic is rate-limited per key and per workspace, per minute and per day, with a separate budget for paid actions; every response says how much budget is left.
• The change feed now records hard deletes, knowledge-source and report activity, uploads and folders, and names who made each change.
• The new-key dialog groups the new scopes — Settings, strategy & reports, Knowledge base, Workspace Operator powers — so a key carries exactly what you meant.
• An agent asking what it may call now sees the served resource families, its powers and its rate limits in one place.

Fixed

• Agent API errors that should have been 400 or 404 no longer surface as generic server errors.
• A content-only agent key could not read back the receipt of its own write; every read scope can now read receipts.
• Archiving a page, board or database from the app now shows up in the agent change feed.

New

• A new "Operate the whole workspace" option lets an agent you trust read and update every page, database and whiteboard in a workspace — not only the ones it created — while still respecting the conflict checks, and never sharing, billing, members or keys.
• An agent is now a durable identity with replaceable keys: issue a new key for the same agent and revoke the old one, and everything it created, was granted, shared or recorded stays with it.
• Every artifact an agent reads now says who created it — you, or which agent — and why the agent may see it: its own work, a grant, or the operator role.
• An agent asking what it may call is now told the truth about every scope a route accepts, so an umbrella or operator key no longer sees capabilities it actually has marked as denied.
• Workspace owners can move artifacts stranded behind a revoked agent identity to its successor with a dry run first and a rollback lever.

Improved

• Anything under the agent API that is not content — keys, members, billing, secrets — is refused outright for every key.
• Agents can read an archived page on request, the same way they already could for databases and whiteboards.
• The connector tab shows one card per agent with each of its keys, when they expire, and per-key revoke and issue-new-key controls.

Fixed

• The "last used" time of an agent in Settings was never updated; it now reflects real use.

New

• Your own AI agents — Claude Code, Codex, OpenClaw and others — can now work inside a workspace with an API key: reading and editing pages, databases and whiteboards through precise, conflict-checked changes instead of full rewrites.
• Ask the assistant to share a page or database. After you press Approve it mints — or reuses — the read-only public or organisation link and hands it back in chat.
• You can grant an outside agent maintenance of a specific page, database or whiteboard it did not create, and revoke it instantly; every grant and revocation stays on an audit trail.
• The new key dialog issues agent keys with granular scopes — read or write per content type, or one umbrella for all workspace content — so a key carries exactly the access you meant.
• An agent that runs into a capability the workspace does not have files a structured feature request automatically — deduplicated, with a counter for how often the gap is hit — instead of apologising in chat.
• The landing page's Solution section is now a clickable mini Workspace OS you can try before signing up.

Improved

• Every Settings tab now shares one layout — a centred column with consistent headers instead of cards nested in cards — and the API keys page became a compact token list.
• Brand Voice and Target Audience in Customize are proper multi-line fields instead of single-line boxes clipping a paragraph.

Fixed

• A page shared publicly could open blank for its readers when it carried live collaborative edits or a home widget; shared links now always show the current document.
• Accepting the assistant's page suggestion failed with a mismatch error on collaboratively edited pages.
• Agent API keys stopped working after the person who created them switched organisations.
• Settings tabs could briefly crash with "This tab couldn't open" while signing in.
• Chat commands like "grant access to this page" were misread as edit requests when a page was focused.

New

• A page can now carry a live, interactive chart or diagram built by a tool you connect. It starts automatically in browsers that support credentialless frames; unsupported browsers keep showing its plain-text description.
• An interactive block runs sealed off from your account: it cannot read your session, your cookies, or anything else on the page, and you can edit its title, code, data and height right on the page.
• X drafts accept a video you upload yourself — MP4 or MOV, up to 512 MB and 140 seconds — instead of only ones generated for you.

Improved

• Your workspaces now nest under the organisation they belong to, in the single menu at the top. The separate workspace switcher is gone from the sidebar, which starts at your files again, and switching organisation no longer closes the menu you were choosing from.
• Settings is tighter throughout: Language is one row instead of a card, a logo you already have sits beside its field rather than under a large dashed box, Agent's three near-empty cards became one, and the tab strip no longer hides itself behind its own scrollbar.
• Tools you connect can now read the exact shape of a page they are allowed to build, so an integration can be written against a published contract instead of guesswork.

Fixed

• A page could go quietly read-only after a sync error — resize handles and column dividers vanished, and content looked lost — while it was still whole on the server. It now repairs itself instead, and a client that cannot read one update can no longer overwrite that content for everyone else.
• A chat stream that a workspace refuses no longer retries forever; it stops and says the thread is unavailable.
• The workspace activity indicator survives a reload, and stops showing runs that already finished or were cancelled.
• Brand colours in Settings no longer overlap their own label.

New

• Drag one block onto the edge of another to put the two side by side. A layout holds up to five columns, and the dividers between them drag to set how wide each one is.
• Galleries are back: type /gallery for up to ten images in one block, switch it to a swipeable carousel, and give any image a title to turn its cell into a card.
• Inserting an image, video, audio or file now offers everything already uploaded to the workspace, so the same picture can appear on any number of pages without being uploaded again.
• Bookmarks, embeds, HTML blocks, files, linked pages, whiteboards and images can all be resized by dragging an edge — or set to a third, half or full width — and aligned left, centre or right.
• A bookmark card now carries the site’s icon and name beside its description and image, and rearranges itself to fit the width you gave it.
• A bookmark, an embed, a mention chip and a plain link in text now convert into one another without losing the address or anything already fetched about it.

Improved

• Galleries and columns now respond to the width of the pane they sit in rather than the width of the window, so a narrow pane stacks them correctly on a wide monitor.
• Pages have narrower side margins on a phone, giving the text itself about 70 more pixels.
• A page you share publicly now opens with just the ActVox mark and a single button, instead of the full product menu — every item of which led your reader to a sign-in wall.

Fixed

• “Turn into” no longer destroys the block you use it on. On an image, bookmark, embed or gallery it offered text conversions that silently deleted the block or rewrote the paragraph beside it; those blocks now offer only conversions that keep their content.
• An image used on several pages no longer disappears everywhere when one of those pages is deleted.
• Asking the assistant to change a page you have open now changes that page, instead of creating a near-duplicate next to it.
• Column widths you set now actually render — layouts previously came out stacked inside a single column whatever ratio you chose.
• Resize and alignment controls appear when you hover a block on a freshly opened page; before, they stayed hidden until you had clicked something.
• A table on a shared page no longer overlaps its own cells on a phone: it scrolls inside its own frame and the page no longer slides sideways.
• A long web address in a shared page no longer stretches the whole document wider than the screen.
• A page whose collaboration history could not be replayed no longer stays blank after being repaired.
• A page created through a connected agent can be edited again even when it holds a table, a gallery or another rich block — and editing one no longer quietly strips its images, sizes or icons.
• On a workspace past its thirtieth page, the assistant can now see the newest ones; before, they were invisible to it and could not be named as a target.

New

• Paste a link and choose what it becomes: an inline mention carrying the site's favicon and real title, a bookmark card, or a live embed. The choice appears beside the link and disappears on your next keystroke, so it is offered without being imposed.
• YouTube, Vimeo, Loom, Figma and Google Maps links now embed from the URL you actually copy — previously the copied form was precisely the one that refused to display.
• Drag a tab to the right or bottom edge of a pane to split the workspace there. Items dragged out of the sidebar work the same way.
• Whiteboards created from a page now sit under that page in the sidebar, and can be dragged into and out of pages. Deleting a page with nested items asks what to do with them.
• Images take a caption, and clicking one opens it full-size with zoom and download.
• A card linking to another page now reads that page live — its current title, icon and preview — and can be resized.
• Claude, Codex and Hermes appear as connectors, each with a setup tab that installs the skill and issues a key scoped to that agent, so one can be revoked without touching the others.

Improved

• A workspace now holds at most seven open tabs, with “Close N other tabs” in the tab menu. Eight or nine tabs left every one of them too narrow to read.
• An HTML block sizes itself to its content instead of cropping charts and diagrams at a fixed height.
• The database table keeps its title column in place while you scroll sideways — on a tablet that is the difference between reading a table and reading rows of anonymous values.
• The slash and @-mention menus open at the caret and flip above it when there is no room below, instead of opening off-screen.
• The marker for a running task is a spinner rather than a sparkle, and it stays on the object being worked on when you switch tabs.
• Your workspace ID now sits in Settings → API keys, beside the keys it is used with.

Fixed

• An image that was still uploading is no longer lost if you reload, and the upload now shows its progress in the page.
• Switching to a fourth tab no longer abandons a content generation that is still running.
• A link preview no longer takes its title from a bot-protection page, and re-checking a link can no longer blank a preview you already had.
• A site that refuses to be embedded now shows a bookmark card instead of the browser’s “refused to connect” page — including in pages where such a block was already sitting.
• Pages no longer begin with a blank line between the title and the first block.
• Connector settings pages scroll; everything below the fold was previously unreachable.

New

• Your workspace now has a permanent API address — https://api.actvox.com — for connecting your own AI agents. Set it once; it keeps working even as we move our infrastructure underneath it.
• Settings → Agent tooling shows the address and your workspace ID together, and the copy-paste setup block uses the permanent one.

Improved

• Affiliates now earn a flat 30% from their first paying referral, for 12 months per referred customer — no tiers to climb.
• Pages that moved to actvox.com now redirect there from the app instead of returning “not found”.

Fixed

• A chat connection that was refused no longer keeps retrying itself in the background for hours.
• Diagrams, tables, checklists and images written by a connected agent stay editable afterwards instead of being rejected on the next edit.

New

• Attach your own files to a content draft: several at once, in the order you choose, each removable on its own.
• Instagram, LinkedIn and Facebook posts carry your uploaded media through to publishing, including multi-image posts.

Improved

• Rating an answer now offers you a note straight away, and asks the right question: what worked well, or what went wrong.
• Your note is easier to find — the button says what it does instead of showing a bare icon — and hovering it shows the whole note.
• Saving a note now confirms it saved, and a save that fails keeps your text instead of discarding it.
• The rating buttons and the note work properly with a keyboard, and read correctly in a screen reader.

Fixed

• Files are checked as you attach them, so a post that would be rejected tells you before you publish rather than after.
• The affiliate signup link no longer points at the wrong page.

New

• Rating an answer can now carry a written note, so you can say what was wrong instead of only that something was.
• Your note is saved with the rating: reopen it later to read it, edit it, or remove it.
• A task can have more than one assignee — search your team, tick everyone responsible, and each name shows on the row.

Improved

• Everyone newly added to a task gets notified, and re-saving the same task no longer stacks duplicate notifications.
• Disliking an answer no longer replies in the chat asking you what was wrong — the note replaces it, and it costs you nothing to leave.

Fixed

• Removing a reaction from an answer now reliably clears it, instead of occasionally leaving the old rating in place behind the scenes.
• Feedback can now carry a written explanation alongside the rating, so we can see what actually went wrong rather than just that something did.

Fixed

• The assistant would sometimes say it could not create a page or database that it had, in fact, just created — it now recognises the item even when the name it saved differs from the words you used.
• When a chat turn failed, the panel could go completely blank with no message at all. The reason is now always shown.
• Members of a team workspace could be locked out of image, video and content generation on their own workspace. Access is now checked against your organization directly, so a session hiccup cannot shut you out.
• Image generation could spin forever instead of reporting a problem, and quietly retry itself in the background.

Fixed

• Generating images or video in a team workspace no longer fails with an access error after your session quietly drops out of the organization.
• A chat tab left open on a workspace you can no longer see now stops cleanly instead of retrying in the background for hours.
• Your organization is remembered correctly across sessions, and membership is now re-checked with your identity provider each day.

New

• Resize any image in a page: drag either edge, use the 33% / 50% / 100% presets, or nudge it with the arrow keys.
• Add a whiteboard straight into a page from the slash menu — it opens in its own tab and nests under the page.
• The sidebar is now a proper tree: arrow keys move through it, and every row has a right-click menu.
• Rename and delete a whiteboard from the keyboard, and reorder pages without a mouse.

Improved

• Attachments moved from a permanent strip at the foot of the page into a header button, so the document starts where you left it.
• Pages open faster — the database grid no longer downloads with every page.
• Chat context loads in parallel, so the first reply comes sooner.
• Switching organizations no longer bounces you through a redirect.

Fixed

• AI content generation failed every time in production — it works again, and generated drafts arrive as expected.
• Pages could quietly stop saving after a connection hiccup: the editor kept accepting text that never reached the server. Saving now recovers on its own.
• A page could open blank and refuse to accept typing until reloaded.
• Mermaid diagrams drew every box empty — labels are back, and the diagram is centred.
• Clicking a diagram, embed, or file block no longer pops the text toolbar over an unrelated part of the page.
• The cookie banner no longer sits above dialogs and makes them unusable on a first visit.
• A skill's "Try it out" prompts now open the chat and fill in the prompt.
• Image generation no longer reports "in progress" forever when it has actually failed.
• Dates now follow the language you chose in the app rather than your browser's.
• Keyboard focus is visible again across the editor, database grid, and comments.
• The block menu closes when you click away from it.
• Reduced-motion settings are now respected by menus, dialogs, and loading states.
• The sidebar reads correctly right-to-left in Arabic.

New

• Connect Google Search Console to a workspace: pick a property and ActVox imports 16 months of search history, then keeps daily clicks, impressions, and positions flowing in automatically.
• Creating a workspace now happens right in the app — one dialog, two paths: start blank for docs and chat, or drop your brand link and get an analyzed, pre-filled workspace.
• Meta Ads and Google Ads show as upcoming connectors — press Request and your vote is counted.

Improved

• A brand-new workspace opens straight into a setup chat that offers to run your report or build your profile with you.

Fixed

• Creating or entering a workspace no longer triggers the browser's "Leave site?" warning.
• Chat could silently stop answering in some team workspaces after a session hiccup — every send failed in the background. It answers again.
• New Content generation no longer stalls forever — runs recover after reconnects, retry on failure, and settle or refund credits exactly once.
• A connection blip or a suspended tab no longer wipes the open workspace — your work stays visible and a recovery action appears.
• Deleting a workspace no longer flashes "Workspace not found" errors from tabs that were still open.
• Turning a report into a workspace could silently skip creating its content drafts; they now generate reliably in the background.
• Edits made in the artifact editor right before closing the tab are no longer lost.

Improved

• Typing in a page no longer slows the rest of the workspace — the sidebar and file tree used to refresh on every keystroke.
• Moving through a database with the keyboard is much faster: an arrow key now redraws a handful of cells instead of the whole table.
• The workspace downloads noticeably less code up front — editors, row cards, and studio tools load the moment you first open them.
• Typing in a content draft stays smooth while an AI edit suggestion is open, and image previews stop re-rendering on every character.
• The workspace shell no longer re-renders on every chat message, and long streamed replies do less background work per token.
• Galleries and page pickers transfer far less data — they no longer download full document bodies or image editing data just to show titles and thumbnails.

New

• Select a passage and press "Improve writing" or "Ask AI" — the agent now edits ONLY that passage, shown as a proposal you accept or reject. It works even when the selection spans several paragraphs.
• Short, precise replacements go through — shrink a selected phrase to a single word and the edit is accepted.

Fixed

• Some agent requests ended with "I could not complete that" even though the work was possible. Two causes of those doomed runs are gone: impossible internal requirements and garbled titles parsed from your request.
• Competitor analysis no longer loses its data when a step times out — partial results are kept, and an empty section shows the reason with retry guidance instead of going blank.
• Web pages and files that captured no content used to sit in your knowledge base as silently empty sources. They now show as failed with the reason, and you can re-index them.

New

• A database row opened as a page now shows all of its properties, editable in place — not just its title and text.
• You can add a new property right from a row's card or page — no need to switch to the Table view.
• Databases in the sidebar now expand to show their rows, so you can jump straight to a row's page.
• A new Sitemap page at actvox.com/sitemap lists every public page in one place, in all four languages. You can reach it from the footer.

Improved

• Chat looks cleaner and matches the rest of the workspace — simpler message rows and composer.
• Sign-in and shared pages load with about a third less code, and heavy views now load only when you open them.
• Chat stays smooth while a long answer streams in, and the command palette no longer lags while you type.

Fixed

• Card modals, delete confirmations, and other dialogs opened off-screen, and menus inside them did not respond to clicks. They now open centered and work.
• Deleting an item could freeze the whole page until reload. It no longer does.
• Renaming or quickly creating select options no longer loses your changes.
• Creating a row while filters are active now opens it as a card instead of appearing to do nothing.
• Deleting a row while its page was open showed repeated "Page not found" errors. It closes cleanly now.
• Databases embedded inside a page no longer fight your clicks while you work in them.
• Linked calendar cards now open the content editor reliably by mouse or keyboard, and dialogs describe themselves to screen readers.

Fixed

• Every page's "My Workspace" link and the Try it links on What's New led to a not-found page instead of your workspace. They now open the workspace correctly.
• The buttons on the Pricing page led to a not-found page. They now open Billing, where you can pick a plan or top up credits.
• Help, Cookie Policy and out-of-credits links inside the workspace led to a not-found page. They now open the right page.
• Video generation would have failed to finish, because the completion callback pointed at the retired research.actvox.com address. Fixed before it affected anyone.

Fixed

• Reports and notifications now arrive from actvox.com instead of the retired research.actvox.com address. If you filter or allowlist ActVox email, update the rule to actvox.com.
• The support link on the upgrade screen reached an address that could not receive mail, so those messages bounced. It now goes to a working support inbox.

Fixed

• Page addresses shared with search engines and social previews now point at actvox.com instead of the retired research.actvox.com.

Fixed

• Opening os.actvox.com directly now takes you to your workspace instead of showing a not-found page.

New

• ActVox now lives at actvox.com, with your workspace at os.actvox.com - the public site and the workspace are separate applications, so each ships on its own schedule.
• Connect Hermes, Codex, or Claude Code from Agent tooling with a verified installer and a workspace-scoped key.
• Document Studio turns your workspace evidence into decision memos, PRDs, research reports, and operating plans.
• External agents can create, update, archive, restore, and optionally share workspace pages through one versioned API.
• Knowledge Graph now groups related nodes into named clusters and answers structural questions without guessing.

Improved

• Agent keys expire after 90 days, stay scoped to one workspace, and cannot publish public links unless you opt in.
• Report requests that could mean a website audit or a written document now ask which one you want.
• Company analysis now shows the real 8–30 minute range and tells you when it is safe to leave.

Fixed

• Links to the old research.actvox.com address no longer work - update bookmarks, and re-send anything you shared from it.
• Everyone is signed out once as part of the move; sign in again at os.actvox.com with the same account. Workspaces, reports and content are untouched.
• Schedule edits made while calendar data loads no longer disappear or leave Save disabled.
• Users with enough credits for a basic report are no longer blocked by the deep-report price.
• If credits are insufficient, analysis now says so instead of looking like a generic failure.
• Reports without a usable marketing strategy no longer show an Apply report action that cannot work.

New

• Whiteboards are a new kind of workspace object - an open canvas beside your pages and databases.
• Two people can draw on the same whiteboard at once, each seeing the other's cursor move.
• Write a diagram as plain text in a page with the Mermaid diagram block and watch it render as you type.
• Ask the assistant for a flowchart or a diagram and it draws one, in a page or directly in chat.
• Turn any diagram into editable shapes on a whiteboard with "Edit visually on a whiteboard".
• Shared pages now render their diagrams for anyone opening the link.
• Tell the assistant to remember a fact, ask it what it remembers, and tell it to forget.
• Select text in a page and choose Improve writing or Ask AI without leaving the editor.
• The Properties panel on the right now inspects the page, database or conversation in front of you.
• A badge on the tab shows which page the assistant is working on while it works.
• Pages and databases the assistant creates now open by themselves instead of waiting to be found.
• Your own tools can create pages, databases, whiteboards and drafts through the ActVox command line.

Improved

• The text selection menu is now a vertical list of labelled actions, anchored beside what you selected.
• Rating an answer with thumbs up or down now feeds into how the assistant improves over time.
• Approving or rejecting the assistant's proposed changes teaches it as well.
• Eight blog articles were rewritten with substantially deeper, more practical content.
• Diagram text is included in search, so a page can be found by what its diagram says.

Fixed

• Deleting a workspace now stops it serving data immediately, rather than only once its contents were erased.
• Work already running in the background can no longer write into a workspace you just deleted.
• The assistant no longer goes quiet when a step fails - it now tells you what went wrong.
• "New whiteboard" now appears in the + menu beside your workspace name.
• Live cursors on a whiteboard work again.
• Asking the assistant to make a whiteboard now reaches the whiteboard tools instead of being misread.
• Repeating a diagram request no longer adds the same diagram to the board twice.
• Accepting a suggested page edit no longer fails with an error over changes you cannot see.
• Searching the Knowledge Graph now repaints the matches instead of restarting the layout.
• The Create API Key dialog stays usable on short and narrow screens.
• A carousel that comes back with fewer slides than asked now says so instead of claiming the full set.

Fixed

• The affiliate terms page no longer advertises translations it doesn't have to search engines.

New

• Upload your logo once - Settings and the Brand hub now share one logo, and every surface in your workspace uses it.
• Ask for an image with your logo and the real logo file is stamped on after generation - never an AI-drawn imitation.
• A new Placement panel in the Brand hub sets where the logo sits on each format - square, portrait, landscape or story - with a live preview.
• The image editor's Logo button now starts from your saved placement instead of a fixed corner.
• A logo discovered by your website analysis is saved into your workspace permanently, so it keeps working even if the original file moves.
• Three new articles on the blog, now with imagery.

Improved

• A report that couldn't reach outside data now carries a clear notice instead of quietly looking complete.

Fixed

• Uploading a logo now lights up the editor's Logo button and reaches the assistant's design context right away.
• Dashboard report tiles show the business's own logo instead of the ActVox logo.
• Chat no longer cites your Knowledge Base when it actually found nothing there.
• Reports no longer invent competitors, sources or statistics when a data source is unreachable - they say what's missing instead.
• The cookie-consent banner no longer covers the language switcher at the bottom of the page.

New

• Everything the assistant writes now follows your brand voice, not just your colours and fonts.
• Write your own voice for each channel in Brand DNA, and drafts sound like you instead of a template.
• Ask the chat to create a report and it starts the real site analysis for you.
• When a request could mean two different things, the assistant asks first instead of guessing.
• Every draft says which brand context it used, so you can see what the assistant was working from.
• The compose wizard warns you before you spend a credit on a workspace with no brand set up.
• Attach your own MP4 or MOV to an Instagram Reel draft and preview it as a Reel.
• Select text on a workspace page and leave a comment exactly where it belongs.
• Commented text stays highlighted, so you can find the conversation again later.
• Mention a workspace member in a comment and they are notified once, not repeatedly.

Improved

• Your channel's writing voice now applies by default, instead of only when you remember to pick it.
• Carousel prompts open ready to edit, and keep your wording once you change them.

Fixed

• Drafts no longer come back in the wrong language when your style guide is written in English.
• Asking for a report in Russian, Ukrainian or Arabic now works instead of being ignored.
• "Create a report" gives you the report you meant, not a blank document.
• The assistant can finally file a chat attachment into your Knowledge Base when you ask it to.
• A file that failed to index says so, instead of showing "Indexing" forever.
• Brand DNA colour, theme and typography pickers use the app's own menus again.
• The compose wizard no longer claims you have no Brand DNA when you do.
• Picking a writing voice right before generating no longer applies the previous one.
• Closed a gap that let a signed-in user read another workspace's writing voice.
• A carousel that hits a rate limit now waits and retries instead of quietly returning fewer slides.
• When a carousel cannot be finished, you get your credits back and a clear partial result.

New

• Right after sign-up, your chat opens mid-conversation: your request, then a live card for the report already being generated.
• Open a report while it is still generating and watch sections fill in as they finish.
• When the report completes, the chat says so on its own - no refresh needed.
• Files you attach in chat become workspace files instead of vanishing when the conversation moves on.
• Artifacts gains an Uploads section with search, filters, and an In Knowledge Base marker on indexed files.
• Ask the agent to file an attached document into your Knowledge Base - or let it offer, and just answer yes.
• Add or remove any upload from the Knowledge Base yourself, from its card menu.
• Attach your own MP4 or MOV to a TikTok draft instead of generating a video.

Improved

• Attaching a file no longer indexes it automatically - indexing is a separate, explicit act you control.

Fixed

• While your report is running, the assistant now says so honestly instead of claiming nothing was started.
• The agent no longer files a document you told it not to file, or re-files one because you asked about a past filing.
• Deleting a knowledge source no longer strands the original file - it returns to your uploads.
• Asking for several pieces of content at once (like a post and a video) no longer makes the assistant go silent.
• The report-ready message in chat arrives reliably and its card shows the right business name.
• Board cards no longer flash back to their old column after you drag them.
• Knowledge Base actions only report success when the file was actually indexed or removed.
• Attaching a long document no longer confuses the assistant about what you actually asked.

New

• Connect GitHub to your workspace and pick up to three repositories for the agent to read.
• The GitHub connection is strictly read-only - it never pushes commits or opens pull requests.
• Refresh a report on demand - each run adds a dated snapshot instead of overwriting the report.
• Turn on a monthly automatic refresh per report; it is off by default and skipped when your balance is too low.
• A History panel beside an open report lists every snapshot and what changed in each.
• Download a complete copy of your workspace - profile, pages, chats, drafts and reports - in one bundle.
• Reports, Strategy and Calendar are their own sidebar destinations again.

Improved

• The sidebar's Vault section is now called Studio, holding Content, Calendar, Brand DNA and Knowledge Base.
• Customize moved to the sidebar footer with the other account controls, and Trash is back down there too.
• A report refresh re-runs only the audit and strategy - it never regenerates your drafts or images.
• The home page headline is sized to fit the screen instead of dominating it.

Fixed

• The verification-code boxes on sign-up and sign-in are visible again - no more typing into blank space.
• Platform content plans no longer time out and ship empty sections, and a plan that fails says so instead of claiming success.
• Starting a report no longer claims you are out of credits when you can afford one.
• The Calendar tab has a proper header and its own create button.
• Finishing a video shows one success notification, even after a page refresh.
• Editing a carousel prompt keeps the generated brief when you switch modes.
• Telegram bots connect against the correct address again, so connected bots receive messages.
• Deleting a workspace or your account now removes everything it owned, including connected social credentials.
• Closed several security gaps around file deletion, file links and GitHub repository access - each now requires proven ownership.

New

• Billing is now a row in your sidebar, showing your credit balance at a glance.
• Opening Billing keeps you in your workspace instead of sending you to a settings page.
• Choose exactly what a page cover shows - drag it around, zoom in, and set how tall the band is.
• Cover framing stays editable, so you can re-crop a cover any time without re-uploading it.
• Links you share now preview with the real page title, description and cover picture.
• The workspace switcher shows each workspace's own logo instead of a letter.

Improved

• Running out of credits now opens Billing right where you are, instead of throwing you out to the pricing page.
• Cover photos are resized before upload, so pages load faster for you and for anyone you share them with.
• Shared pages are readable on a phone - bigger text, tighter margins, and a cover that no longer fills the screen.
• Private shared links stay private in previews: nothing from a workspace-only page leaks into the preview card.
• Green accents across the workspace are now reserved for status, so they mean something when you see them.

Fixed

• Content generation finishes instead of reporting a timeout on work that had actually succeeded.
• Generating for one platform now costs one platform - picking a single network no longer runs all four.
• A platform that fails during generation says which one it was and why, instead of showing an empty error.
• Social posts are no longer rewritten a second time for missing sources they were never meant to cite.
• Telling the assistant its answer missed the mark now saves a single follow-up question that survives a reload.
• Dragging a page cover tracks your cursor exactly, instead of racing ahead of it.
• Replacing a cover no longer drops the new picture at the old one's crop.
• A failed cover upload now tells you what went wrong instead of silently doing nothing.

New

• Report sections now open as proper forms - ratings, review counts, categories and issue lists each get their own field.
• Your strategy's Action Plan and Implementation Roadmap edit as fields too, including the completion checkbox that was previously invisible.
• A bar at the top of a document counts your unsaved changes and lets you save or discard them all at once.
• Artifacts show what they actually are - presentation, document, spreadsheet, image or video - instead of labelling everything the same way.
• Image and video artifacts show a real thumbnail on their card.
• Sort artifacts by most recent, newest, oldest or name, and switch the whole page between cards and a compact list.
• Delete a skill straight from its page - no need to open the editor first.
• The skill editor tells you which skill a test message would actually reach, so you catch trigger-phrase clashes before saving.
• See a live preview of your skill's card while you write it, and add the “Try it out” prompts that appear on its page.
• Typing /help now lists what your assistant can genuinely do, lane by lane, plus an honest list of what it cannot.

Improved

• ACTVOX.md, DESIGN.md, Strategy and your reports sit at the top of Artifacts as clear, colour-coded cards.
• Your assistant now follows the standing rules in ACTVOX.md when it creates and changes things, not just when it reads.
• Your assistant is better at knowing its own abilities, so it is far less likely to claim it cannot do something it can.
• The skill editor drops the “Parameters JSON” box that never did anything, and explains what each remaining field is for.

Fixed

• The Strategy tab no longer shows raw code-like labels on its fields/JSON switch.
• Report editing is fully translated - Russian, Ukrainian and Arabic readers no longer get an English-only form.
• The editing toolbar no longer hangs outside the text column, and switching between viewing and editing no longer shifts the page.
• Saving an action plan no longer fails outright because of one unexpected value.
• Typing a character and deleting it no longer leaves a document convinced you have unsaved work.
• The unsaved-changes bar now clears once your work is saved, instead of claiming changes that already landed.
• The Artifacts header no longer reports zero artifacts above visible cards.
• The sort control is reachable in a workspace that has documents but no artifacts yet.
• Opening the skill editor no longer logs an accessibility warning for screen-reader users.

Improved

• Longer video scenarios get a generation-time budget that scales with their length instead of a fixed cutoff.

Fixed

• Long-form videos (16–120 seconds) now complete reliably - they previously failed right after the first segment.
• Videos no longer get stuck on “generating” forever - an automatic recovery check now finds any finished video whose completion signal went missing and delivers it within minutes.
• Seven videos that had been stuck from previous weeks were recovered and delivered to their drafts - no credits were lost.
• A brief provider hiccup can no longer cancel and refund a video that is still generating.
• A delivered video can no longer be overwritten or marked failed by a late duplicate update from the video provider.
• Opening the app no longer fires a burst of notifications for videos completed long ago.

New

• Pin chats, pages, databases and generated artifacts to a Pinned section at the top of the sidebar, and drag to reorder them.
• Recent in the sidebar now covers everything - chats, pages, databases, artifacts and running agent tasks - sorted by last activity.
• Filter Recent by type, status or activity window, and group it by date, type, status, parent page or unread - your choices stick per workspace.
• Collapse any Recent group with a click; a collapsed header shows how many rows it hides.
• Archive a chat from its row menu, and an unread dot marks conversations that moved since you last opened them.
• Skills and Connectors - one page to browse your assistant's skills, read what each one does, and try curated example prompts.
• Connect LinkedIn, Instagram and Telegram from the same page, and request an upcoming connector with one click.
• The chat composer's new + menu attaches files, recent items, project pages and skills - no need to type @ and know the name.
• Settings › Customize controls which sidebar entries show, the order of sections, the Recent page size, and your business profile.
• Analytics is now a Settings section you can link to directly.
• ACTVOX.md, DESIGN.md, Strategy and your reports live as document cards at the top of the Artifacts tab.

Improved

• The sidebar's five sections became three - Vault, Workspace and Recent - with Search, New chat and Artifacts one click from the top.
• Expanding one sidebar section no longer collapses the others, and reopening the app restores every section you left open.
• Three brand rows that all opened the same place merged into one Brand DNA entry.
• Clicking a selected Recent filter clears it, and clicking a type isolates it instead of removing just that one.
• Account connections moved from Settings to the Connectors page; the old Settings link still gets you there.
• Connector logos render in their real brand colours and ship with the app instead of loading from a third party.

Fixed

• Analytics could show numbers from a workspace you had already left - it now always reports the one you are looking at.
• Archived chats no longer resurface in every list; find them under Recent's Status filter.
• Two links in Settings reloaded the whole app instead of switching sections.
• Guided-tour steps that pointed at moved or renamed buttons anchor correctly again.
• Pinned rows align correctly in Arabic.

New

• Export to Excel from a table, a chat answer or a report - numbers and dates arrive as real values.
• Ask the assistant for a table "in Excel" and get a workbook that always reflects current data.
• New database dialog with marketing templates, several pre-filled from your own reports, calendar and strategy.
• Duplicate an existing database's structure - every column and view setting, none of the rows.
• Resize table columns and rows by dragging their edges; double-click either one to fit the content.
• Set alignment per column, and every table remembers whether you left it on Table, Board, Calendar or Gallery.
• An in-app date picker, plus type-to-create options in Status and Select cells.
• Knowledge Graph - an interactive map of your sources and how they connect, with search and live build progress.
• Artifacts tab - browse everything the assistant generated, with search, filters and a Trash you can restore from.
• Knowledge Base tab - your indexed documents as a library with status, chunk counts and one-click retry.
• Reports tab - every analysis as a card carrying its score, industry and status.
• What's New page explains the handful of changes worth knowing about, in your language.
• Drag nested pages and databases to reorder them, or move them with Alt and the arrow keys.
• Strategy View shows your Brand DNA - brand mark, palette, typography, values and positioning.

Improved

• Switching workspace tabs keeps the last three mounted, so drafts, cursor, scroll and undo history all survive.
• Page headers follow a Notion-style layout: a larger icon over the cover and one hover-revealed row of actions.
• Nested pages and databases render as inline links you can drag into a pane, into chat, or onto a folder.
• Knowledge, Artifacts and Reports open in one click from the sidebar instead of unfolding a tree.
• Changing a column's type now tells you how many values will convert and how many will clear, before it runs.
• Card fields use the app's own dropdowns and calendar instead of grey browser widgets.
• Export headers follow your workspace language rather than always being English.
• Shared page links show the same cover, icon and title as the app.
• The browser tab icon mirrors the emoji of the page you are reading.
• The breadcrumb ends with the page you are on instead of a trailing slash.

Fixed

• Page edits are no longer lost when a save is rejected - they retry, and a failure now tells you.
• Half-typed text survives leaving a tab: chat messages, table cells, renames, comments, captions and brand fields.
• Table edits land on the row you were editing, even when a sort moves it out from under you.
• Status, Select, Date and Assignee cells open their editor when you click them - before, nothing happened.
• Every cell type can be edited with a mouse: one click selects, the next opens the editor.
• Changing a column's type converts the values inside it instead of leaving unreachable data behind.
• Cells no longer shrink or jump when you start typing, and text cells grow as you write.
• Page titles no longer render invisible after a tab switch.
• A dialog left open in a background tab can no longer freeze the whole app.
• Keyboard shortcuts only act on the tab you can see, and background tabs stop playing audio.
• Renaming Home now shows up in the sidebar, and tables embedded in a page can create rows in place.
• Column headers show their type, so a renamed column is no longer mistaken for a broken one.
• The assistant no longer invents a page body when you only asked for a title.
• Report share menus stay inside the pane, and Instagram previews open the image lightbox again.
• Artifacts could be read or overwritten from another workspace - now they cannot.
• Building the knowledge graph shows your sources instead of an empty canvas when nothing links yet.
• The changelog no longer shows machine-padded stand-ins in place of real translations.

New

• Media Studio - generate image, carousel, and video in a dedicated workspace tab with phone-frame previews.
• Grok Imagine Video 1.5 - text-to-video, image-to-video, reference-to-video, and long scenarios up to 120 seconds.
• Generated videos save onto the draft, appear in the Content Gallery, and open with Watch / Edit video.
• Carousel generation - 2–10 cohesive slides for Instagram, with slide edit and detach tools.
• Brand DNA Hub - scan your website or Instagram, store visual identity, and mark brand files.
• Reference Manifest - add competitor or inspiration sites so generation can respect their visual cues.
• Edit Image after generation - Fabric canvas with logo, fonts, palette, text, shapes, and undo.
• Compose tab - create content by type from the Content tab or calendar without leaving the workspace.
• Content generation uses Brand DNA, strategy pillars, and higher-quality prompts by default.

Improved

• Content Pillars and Platform Priorities edit as structured fields instead of raw JSON.
• Workspace sidebar groups Workspace, Context, Brand, and Outputs more clearly.
• Smarter image and carousel auto-prompts from Brand DNA and visual identity briefs.
• Videos stay playable after provider URLs expire - we store a durable copy.
• Instagram-only brand scan works without forcing a website URL.
• Video generation modal uses the approved green system colors and real keyboard focus on concept cards.

Fixed

• Image editor no longer crashes when the canvas remounts during load.
• Content Gallery shows videos, not only still images.
• Strategy field saves no longer fail silently; every block save reports its outcome.
• Turning a full list into a to-do list keeps every item, not only the first.
• Narrow content editor headers stay usable; icon-only actions show hover tips.
• Media Studio previews no longer clip or show planning briefs as fake captions.
• Inline table resize and controls stay stable in the page editor.
• First-run profile confirmation, knowledge-base recovery, and chat citations behave more reliably.

New

• Four new blog articles - audit-to-workspace workflows, agency SEO/AEO audits, AI content systems, and AI visibility audits.
• Every existing blog post now has a full article body instead of a short summary.
• Thumbs-up and thumbs-down on chat answers are now delivered reliably, surviving reloads and connection drops.

Improved

• A research run that loses part of its data is now marked degraded instead of quietly looking complete.
• Our quality monitoring keeps your prompts, sources, and report content out of trace logs by design.
• Blog articles carry proper structured data, and the footer now links to the pricing and roles pages.

Fixed

• Fixed several search-engine indexing problems: permanently cached language redirects, image and text endpoints indexed as pages, and public links pointing into signed-in areas.

New

• New Team plan at $25 per seat - seats follow your member count automatically, and the token allowance scales with them.
• Your token balance now shows how much you have spent today, right under the balance itself.

Improved

• Workspace, report and member limits no longer depend on your plan - your token balance is the only thing that decides how much you can do.
• Creating a workspace, adding a report and adding a teammate stop asking you to upgrade for reasons unrelated to what anything costs.
• Running out of tokens is now the single clear reason anything is blocked, instead of one of several different walls.

Fixed

• Buying a plan we had just added could take your money and grant nothing, with no visible trace. Any unrecognised plan now records the problem instead of failing silently.
• The analyse dialog refused to start a report for anyone holding between 1,000 and 1,499 tokens, quoting a price from before the token switch - a report costs 1,000.

New

• Billing now runs on tokens - you pay what a request actually costs, not a flat credit.
• New Student and Ultra plans, plus token top-up packs for one-time boosts.
• Every report now shows as an ATX Report dashboard - tabs, score rings, and key-insight tiles.
• Audit findings are graded by severity, filterable, and show their evidence.
• "Apply report" converts a report into your workspace in one click - name, profile, knowledge, and drafts.
• Signing up while your report generates drops you straight into a workspace conversation about it.
• Short guided tours replace the long welcome walkthrough, shown on the real interface.
• Every content draft now has a shareable link.

Improved

• Content, calendar, settings, and reports all live inside the workspace now - old links redirect automatically.
• The report-generating page shows live pipeline progress with a progress bar and next steps.
• Workspace limits are now flat - 5 personal and 10 team workspaces on every plan.
• Knowledge-base indexing got roughly three times cheaper.
• Applied reports carry a visible "Added to workspace" mark, and each workspace keeps one business.
• Social, competitor, and local sections show richer cards with real platform identity.
• A simpler sidebar - setup pages grouped together, feedback in one place.
• Long conversations are billed at cached rates, so extended chats cost less.

Fixed

• A cleanup job no longer cancels report runs that are still working.
• Reports gather sources more reliably and retry when a website scan hiccups.
• The automatic report summary after sign-up is never billed to you.
• Retrying a failed report no longer leads to a deleted page.
• Converting a report into a new workspace now takes you to that workspace.
• Sample posts show their text even when the image is still missing.
• Settings and content tabs keep the address bar in sync as you navigate.

Fixed

• Assistant conversations now record which version produced them, for faster support.

New

• The gallery is a picture wall - images keep their real shape instead of a square crop.
• Every picture says whether it is in a post or unused.
• An unused picture can be placed into a draft that has no image yet.
• Opening a picture shows the prompt that made it, with a copy button.
• You can rate a picture thumbs up or down, and clear the rating.
• You can comment on any picture, with the author's photo and the date.
• Strategy opens as a dashboard - summary tiles, cards and an implementation roadmap.
• Your uploaded workspace logo now appears in the app itself.
• Settings opens as a tab inside the workspace instead of a separate page.
• Any organization's settings open straight from the switcher, not just the active one.

Improved

• One top bar now runs across every workspace page, and it no longer shifts when a sidebar closes.
• Switching organization keeps the page alive instead of reloading it.
• A collapsed sidebar disappears completely instead of leaving a grey strip.
• Sidebars reopen at the width you set them to, and both open at the same size.
• Sidebars no longer rebuild their contents every time you open them.
• Artifacts and Conversations show the latest five with a "Show more" row.
• Content cards reflow to the pane you are working in, not the browser window.
• Typing in chat no longer redraws the tab strip - the workspace feels smoother.
• The switcher now separates Personal from your organizations.
• Help Center lives in the account menu, and "Ask support agent" is now just "Support".

Fixed

• The workspace agent no longer runs on public pages, where it had nowhere to put its work.
• Switching organization no longer leaves a chat retrying in the background for hours.
• Opening a chat that belongs to another organization no longer crashes the page.
• The report tab showed two headers, and its Edit mode could not be edited.
• "Add to a post" opened behind the image viewer and looked like it did nothing.
• Gallery captions printed the entire prompt instead of two lines.
• A page whose address merely contained "/os" lost all of its navigation.
• Strategy's green accents rendered invisible on dashboard pages.
• A conversation carrying an attachment could fail when it resumed.

New

• Home greets you by name with an ask-anything box that starts a conversation.
• Home opens on a real cover image instead of a blank page - and you can swap it.
• The six quick-start actions now sit above the fold, right under your numbers.

Improved

• The dark rail beside the sidebar is gone - workspace navigation is one light panel.
• Switching organization, your account, and Support now live in the sidebar itself.
• Reports, content, settings and calendar pages gained a top bar with navigation.
• Sidebar icons, spacing and collapsed sections now read as one consistent set.
• The Home number row shows visibility score, content published, and credits.
• Home loads faster - its widgets no longer pull your whole calendar to count a week.

Fixed

• Switching workspace or organization no longer raises a "Leave site?" pop-up.
• The Home dashboard shows real numbers again instead of dashes and blank cards.
• Home can no longer be deleted, so your dashboard cannot be destroyed by accident.
• "My tasks" now lists tasks assigned to you, and each row opens the task.
• Knowledge sources no longer go missing when a report adds many of them at once.
• Re-scanning a social profile you already added now works instead of failing.
• Reports index once instead of repeatedly, so search results appear reliably.
• Your credits, tasks and requests no longer blank out while switching organization.
• The feature requests page is now behind sign-in like every other workspace page.
• The Home calendar strip counts the right days in time zones ahead of UTC.
• The workspace still loads when your browser blocks local storage.

New

• Click any image, video, or PDF in chat to open a full media viewer with zoom, pan, and arrow-key navigation.
• Content drafts and settings now open as tabs you can keep open side by side, not full-screen overlays.
• Analytics and Feature Requests open as tabs too - the separate app launcher is gone.
• The assistant can now write and maintain your ACTVOX.md and DESIGN.md, with every change shown as a reviewable diff.
• Attach a reference image in chat and the assistant can actually see it.
• GPT Image 2 joins Grok, Gemini, and FLUX as an image model you can pick.
• Facebook is now scraped alongside the other five platforms, completing your social audit.
• Blog drafts gained a cover image and an Edit / Preview toggle that shows the composed post.
• Tables in a page can be edited - add and remove rows and columns, and drag to resize.

Improved

• The workspace sidebar is on one consistent rhythm, and tabs are shaped like browser tabs.
• Hovering a content card now names where each action takes you instead of hiding the thumbnail.
• The blog editor uses the full width of the panel instead of a small scrolling box.
• Workspace settings and built-in skill names are fully translated in all four languages.

Fixed

• Reports, strategies, blog posts, and content are now written in your language, not English.
• A language picked before you signed up is now remembered.
• Image generation works again - FLUX returned nothing, Grok images were generated then discarded, and the button did nothing inside a draft.
• Generated images keep their real shape instead of being cropped into a thin strip.
• Blog drafts written by the assistant show formatted text, not raw markdown symbols.
• Social and competitor sections no longer arrive silently empty when a scrape fails.
• "Write with AI" no longer fails on drafts containing a footnote.
• The assistant no longer invents the date it collected its research.
• Requests can no longer hang forever waiting on a model that stopped responding.
• TikTok drafts import as the actual script, not a brief or a half-written outline.
• Dragging the divider between panes no longer breaks resizing.
• A finished video now plays in the content preview instead of showing a still image.
• Content freshness is scored against your real content history again.
• Your blog post's cover image is now shown, and gallery images can be previewed.
• The sidebar lays out correctly in right-to-left languages.

New

• Video generation now offers Grok Imagine alongside Kling, with a quality setting of 480p, 720p, or 1080p.
• Ask the assistant for a video and it can pick the model and quality for you.
• Delete a report straight from its ⋯ menu in the sidebar, like pages and databases.
• Database property menus gained Hide, Duplicate, Insert left, and Insert right - in all four languages.

Improved

• Confirmations and name prompts now use the product's own dialog instead of the browser's, so they are styled and properly translated.
• Renaming a database property saves with Enter and reverts with Escape.
• Analytics now wait for your consent, and session recordings mask what you type.

Fixed

• Switching workspace or organization no longer asks "Reload site? Changes you made may not be saved."
• Your pages, comments, and databases no longer flash empty while switching workspace or organization.
• Database edits that fail now tell you why instead of silently snapping back to the old value.
• "Reformat" now works on assistant-written pages that contain bold text or links.
• Video script generation now reports the real reason when it produces nothing.

Fixed

• Pages the assistant created that opened blank now repair themselves - the previous release shipped the repair but it never triggered.

Fixed

• Pages the assistant created no longer open blank - and pages already affected repair themselves the next time you open them.
• Video generation no longer hangs forever when you write your own prompt instead of picking a script.
• Conversations open in the main area again instead of the side panel.

New

• Ask the assistant about your analysis report and it can now read it - competitors, personas, action plan, and every other section.
• Ask for a slide deck and you get one: a real presentation you can scroll, present, and download.
• Databases now work like a spreadsheet - edit cells in place, add formula columns, and sort and filter as you go.
• Import and export any database as CSV, including from the assistant.
• Hide columns you do not need. Hidden columns still power your formulas, filters, and sorts.
• Pick a property type from a searchable list instead of a plain dropdown, and add databases straight from a page with /db.
• Every sidebar section now explains itself on hover, in all four languages.
• Your DESIGN.md and brand colors now shape the images the assistant generates.

Improved

• Databases open as a table instead of a board, so a database looks like a database.
• The assistant no longer claims it cannot write to a database - it just fills it.
• Content generation now has room to finish instead of being cut off partway.
• Long chat sessions stay alive instead of quietly stopping after about 25 minutes.

Fixed

• Fixed a two-week outage where signing in broke content generation, report starts, and image and video generation.
• Pages the assistant writes now render as real headings, tables, and lists instead of raw Markdown symbols. Pages damaged earlier can be repaired with one click.
• "Create a presentation" used to describe a deck in chat and create nothing. It now creates the deck.
• Asking for a database "with the same info" now produces rows, not an empty set of columns.
• Fixed a share-link leak where columns you had hidden were still published to anyone with the link.
• You can now add a row directly in Table view.
• Adopting the strategy from your report is reachable again for everyone, not just a few accounts.
• Fixed a crash that could take down the site on certain asset requests.
• Markdown files the assistant creates now preview properly instead of as a numbered code listing.

New

• Agent approval requests now expire after 24 hours, so stale requests clean themselves up.

Improved

• The workspace loads about a quarter less code up front - pages, databases, and settings now load the moment you open them.
• Chat streaming is smoother: long conversations and open editors no longer redraw on every word.
• Agent replies start sooner after you send a message.
• The command palette (⌘K) and settings open instantly without weighing down the initial load.

Fixed

• Fixed duplicate assistant replies - the same answer no longer appears twice until you reload.
• Fixed a phantom error shown after agent runs that had actually finished successfully.
• Video generation can no longer charge you twice when a retry happens behind the scenes.

New

• Each workspace now opens on a customizable Home page with live widgets for reports, tasks, calendar, and quick starts.
• You can dock a page beside the conversation in a split view, and drag any tab - including chats - between the two panes.
• Tabs are now color-coded by type, so chats, pages, databases, reports, and media are easy to tell apart.
• The assistant can now answer product how-to questions from the Help Center and link you to the exact guide.
• Page cover images can be repositioned and resized to set the focal point and banner height.

Improved

• Pages are full-width by default, with a one-click option to switch back to a narrower reading column.
• New conversations now open as center tabs instead of taking over the side panel.
• Chat titles are shorter and clearer, drawn from what you actually asked.
• The homepage and pricing page have a refreshed, cleaner look.
• The sidebar footer is reorganized into clearer Notifications, Help, Feedback, and Trash rows.

Fixed

• Switching organizations no longer crashes the app with a "Something went wrong" screen.
• The assistant no longer fails a request after it has already created the pages you asked for.
• A wide reliability sweep keeps screens stable while your access context is changing.
• Renaming a page now updates the sidebar as you type instead of only after you finish.

New

• A public Help Center is now available with verified guides in English, Russian, Ukrainian, and Arabic.
• Help guidance now appears directly inside the product where you need it.
• Public pages now live on language-prefixed URLs, so links keep your language.
• Chat answers can now be turned into workspace pages with durable message actions.

Improved

• The homepage opens with a new photographic hero and clearer start paths.
• The website path now says it accepts Instagram, TikTok, and YouTube links too.
• Russian, Ukrainian, and Arabic translations are complete across content views.

Fixed

• The assistant no longer fails after successfully creating pages when your request describes the page instead of naming it.
• Chat error messages now tell you what went wrong instead of a generic apology.
• Page edits, reviews, and creations now save exactly once - no duplicates, lost edits, or phantom content from stale tabs.
• The agent verifies pages were really written before claiming success, in English and Russian alike.
• Onboarding now completes reliably instead of leaving a dead-end screen.

Fixed

• The previous workspace controls are now inert while personal or organization context is changing, including organization-to-organization transitions.
• Failed tenant activation and invitation acceptance now provide visible localized feedback instead of failing only in browser logs.
• Destructive profile replacement identifies its target by workspace name rather than the separate business-profile name.

Fixed

• Switching between personal and organization profiles now reloads through the authenticated workspace resolver, so open documents and workspace-scoped providers cannot survive under the wrong access context.
• Renaming a workspace now updates project navigation consistently instead of being hidden behind the separate business-profile name.

Fixed

• Public agent-context validation now decodes URL paths to a stable value instead of stopping after a fixed number of passes, closing deeply encoded private-route bypasses.

New

• AI agents can read /llms.txt for a concise product map or /llms-full.txt for the complete Workspace OS manual, including onboarding, workspace objects, agent behavior, approvals, persistence, integrations, and current limitations.

Improved

• The public agent manual stays synchronized with the released product version and links only to canonical public ActVox pages.
• Authenticated chat-first onboarding is now documented explicitly as production behavior, while unauthenticated guest onboarding is clearly excluded.
• Search-oriented crawlers and training crawlers are evaluated separately so publisher policy is not confused with search visibility.

Fixed

• Website audits no longer present llms.txt as a ranking requirement or treat its absence as a critical SEO defect.

Fixed

• OpenGraph and Twitter metadata now include an accessible description for the social preview image instead of publishing the image URL without alt text.

Fixed

• Link previews no longer show the retired ATX-OS internal name; their image and accessible description now match the product identity used across the site.

Improved

• Public pages now describe ActVox consistently as an agentic marketing Workspace OS that turns business context into plans and execution, with website audits available as an optional starting point.
• Roles, use cases, blog pages, and social previews now use the same product identity and production domain.

Fixed

• Search engines and audit crawlers can now discover every public marketing page and article through robots.txt and sitemap.xml.
• Canonical links, OpenGraph and Twitter previews, and structured data now point consistently to research.actvox.com instead of the retired development domain.

New

• Start by chatting: paste your website URL in chat to kick off a full report, or begin with no site and the assistant helps from what you tell it.
• The workspace assistant now works in Plan and Act steps and can quote exact text from your pages as evidence.
• Adapt turns any draft into platform-native versions for up to three platforms (X, Instagram, LinkedIn, TikTok) in one click.
• Optional auto-adapt creates platform versions of every new post as ready-to-review drafts.
• Redeem partner promo codes on the pricing page for credits or startup-plan access.
• New startup plan with pro-level limits.

Improved

• Credits are now a clear unit of account with transparent per-action pricing, and your first 10 chat messages each day are free.
• Your report's Marketing Strategy and Personas stay with the report until you explicitly choose "Use this strategy" - no surprise duplicates in your workspace.
• New organizations start on the free plan with 10 credits, and the pricing page shows real pay-as-you-go prices.

Fixed

• Paid credit purchases can no longer be silently lost - grants are processed reliably with a full audit trail.
• Canceling a subscription now correctly downgrades your plan.
• Content drafts with exact titles or specific formats now save reliably as the right type instead of coming back as plain text.

Improved

• Report analysis no longer wastes time trying to scrape login-walled social sites the slow way - it goes straight to the dedicated social scrapers.
• Removed a large amount of unused integration code, making the app lighter and easier to maintain.

Fixed

• If your sign-in session expired in an old tab, the app showed a confusing error - it now quietly asks you to sign in again.

Fixed

• Completion emails were silently rejected by the email provider (unverified sending domain) - the domain is verified and delivery works for signed-in users and captured emails alike.
• Setting a report's notification email now requires owning the report or the original browser session - it was previously open to anyone with the link.

Fixed

• The report's Social Presence section hid scraped analysis behind a "Connect Your Social Accounts" wall - scraped data now renders with confidence badges, and connecting stays optional for verified metrics.
• Social audits now include the real follower counts and engagement the scrapers collect, instead of dropping them.
• Report progress updates write more efficiently behind the scenes (fewer database write conflicts).

Fixed

• The assistant's streaming reply disappeared for a moment (sometimes seconds) at the end of every run, and slow saves could replace a finished answer with an error - the reply now stays visible straight through.

Fixed

• Cross-Platform Strategy rendered as an empty calendar because the per-platform content writers were silently cut off mid-work - all report steps now get realistic time budgets.
• Sign up through a report and your workspace picks up the business context from it - the first-run profile form only appears when you start without a report.

Fixed

• Report analysis lost its website-discovery step to a silent timeout, leaving social media analysis empty and other sections thin - discovery now finishes with whatever it has gathered instead of losing everything.
• Workspace OS icons switched to clean monochrome outline glyphs, with a “Remove icon” option in the emoji picker.

Fixed

• Filename requests ("create Design.md", "fill CONTEXT.md", "заполни DESIGN.md") now reach the right workspace action - create, fill, rename, or delete - instead of a web-search answer.

New

• The agent proposes edits instead of overwriting: review page, draft, and database changes with Accept/Reject diffs.
• Choose how the agent works each turn: Ask (answer only), Plan (approve before acting), or Agent.
• Generate images and TikTok videos straight from chat, with a live progress card.
• Ask for just a picture and get a standalone image - no post attached.
• Pick your image model - Auto, Gemini 3 Pro, Grok, or FLUX - and review four variants.
• One Notion-style Workspace tree: nest databases inside pages and drag to reorganize.
• The agent can now delete, rename, move, and reshape pages, databases, and content from chat - including databases with custom columns.
• Drag any sidebar resource - reports, databases, files, media - into chat to attach it as context.
• Telegram, Instagram, and LinkedIn now connect per workspace, each with its own accounts.
• Connect external coding agents like Codex and Claude Code via workspace API keys.

Improved

• Replies now stream word by word, including the agent's reasoning.
• Sidebar reorganized into Workspace, Context, Agents, and Outputs, with collapsible sections.
• Content the agent writes sounds more natural - shared style rules ban AI clichés.
• A refreshed homepage shows the Workspace OS and both ways to start.
• Page comments now collapse out of the way until you need them.

Fixed

• Working in two windows no longer loses tabs created in the other window.
• A closed split pane stays closed after reload.
• Knowledge Base source discovery now actually indexes the sources it finds.
• Teammates can save and delete shared content drafts without errors.
• Stopping a streaming answer keeps it stopped - no reappearing replies after reload.
• The Telegram bot now connects and replies reliably, in your language.
• Saying "create it" after the agent proposes a document now actually creates it.
• Attached databases in chat now include their rows, so the agent can read them.
• Image generation charges credits exactly once and refunds on failure.
• Security hardening: internal callbacks and workspace API keys now fail closed.

New

• Chat now lives in a panel on the right of your workspace - open any page beside the conversation.
• Switch conversations in place: click the panel title for recent history, or open a thread as a tab or in split view from the conversation menu.
• Send a voice message and the agent transcribes it and answers what you said.

Improved

• Assistant replies use the full chat width - lists and tables are easier to read.
• The agent's reasoning notes are readable now, with calmer styling.
• Cleaner workspace chrome: no overlapping support button, no duplicate account icon, and a visible collapse control on the right panel.

Fixed

• Approving an agent action no longer fails when the action takes more than a few seconds.
• The chat composer no longer gets cut off at the bottom of the window.
• Working in two conversations at once no longer interrupts a streaming reply.
• A deleted conversation can't reappear in the panel, and a cleared panel stays cleared after reload.
• Voice recordings keep their player after a refresh instead of breaking.

Fixed

• Reopening an older conversation no longer duplicates the assistant's answer - the original message keeps its place and gains the reasoning view.
• Conversations already affected are cleaned up automatically, keeping the original timestamps and order.

New

• The assistant can now turn an approved four-week content plan into draft posts and calendar entries in one step - always drafts, never auto-published, and safe to re-run without duplicates.

Improved

• The assistant draws on your workspace memory, instructions, skills, and profile through one consistent, properly scoped channel.
• Upgraded the agent server runtime for faster, more reliable long-running work.

Fixed

• Private assistant memory is now visible only to you - workspace-shared memory stays shared with the team.
• A chat can no longer touch another workspace's data: every request is verified against the workspace you're actually signed into.
• Switching between chats no longer shows a reply from the wrong conversation.
• Assistant replies no longer duplicate or lose their ending, and a finished answer can't be overwritten by an older copy.
• New replies no longer carry over citations, suggestions, or approval cards from a previous turn, and skill results render as clean text instead of raw code.

New

• New ACTVOX.md page in every workspace: write your rules once (tone, language, do/don'ts) and the assistant follows them in every chat and draft.
• Anything you saved under Custom Instructions in settings moves into ACTVOX.md automatically.
• Create your own slash commands - reusable prompt shortcuts with arguments that anyone in the workspace can run.
• New /context command shows exactly what the assistant sees: your rules, business profile, memory, and attachments.
• Every report now ends in one clear action to continue with the agent in a project, on every report page.

Improved

• The / command menu shows argument hints, keeps your text when you press Escape, and ranks the best match first.
• /onboard (or just typing "onboard") starts a real profile interview - one question at a time, saving as you answer.

Fixed

• The assistant no longer deflects normal messages to an onboarding tip - every message gets a real answer again.
• A report that fails late no longer shows "generating" forever, and content generation failures now tell you instead of failing silently.
• Tightened who can view or edit a workspace's commands, and protected the ACTVOX.md page from deletion or renaming.

New

• Paste an Instagram, TikTok, YouTube, X, or LinkedIn account as your starting point and get a full analysis of that profile, not just a website.
• Social reports now surface your content pillars, hook patterns, posting cadence, and top-performing posts, and find rival accounts on the same platform.
• New Feature Requests board: submit ideas with files or a voice note, or send one straight to your workspace assistant to work on now.
• New Help & Support assistant that explains ActVox and how to use it, opened from anywhere.
• The assistant now replies in your language (English, Russian, Ukrainian, or Arabic).
• Cookie and privacy controls with an Accept/Reject banner, per-category settings, and new Privacy, Cookies, and Terms pages.
• In-app notifications now open the right place when clicked, and you get a notification when a report finishes or the assistant replies while you're away.

Improved

• The chat and content assistants now run on a newer, faster model with built-in web and X search for fresher answers.
• Cleaner chat: shorter conversation titles, a calmer composer, and easier access to your chat history.

Fixed

• Security hardening pass: tightened who can access video jobs, connected accounts, and the analysis start flow, patched a sign-in route-protection issue, and sanitized how chat and report content is displayed.
• Analytics no longer runs until you've made a cookie choice.
• A social account with no website is no longer scored as if the profile were a website.

Improved

• Clicking into the chat box no longer draws a hard frame or glowing ring around it.
• Opening a database card now shows its title above the properties, matching the rest of the editor.
• The Strategy item in the workspace sidebar now matches the style of the other top-level nav items.

Fixed

• Ask the assistant to find your competitors and it now reads the competitors already saved in your workspace instead of replying "evidence missing".

Fixed

• Fixed a build-packaging issue that broke plain Enter (new paragraph) in the page and card editors on the deployed site - the previous release's fix didn't take effect in the production build.

Fixed

• Pressing Enter now reliably starts a new line in page and card editors - no more needing Shift+Enter.
• Ask the assistant to find your competitors' social profiles and it now uses the competitors already in your workspace, then pulls their live LinkedIn and social metrics instead of asking you for a list.

New

• A full block menu for pages: headings, lists, to-dos, toggles, callouts, quotes, tables, code, dividers, equations, and more from "/" or the + button.
• Split any page into 2, 3, 4, or 5 columns that stack neatly on narrow screens.
• Drop in live previews - embed or bookmark a URL, link another page as a card, or show a linked database inline.
• Add editable HTML blocks that render as a safe, sandboxed live preview right in the page.

Fixed

• Pressing Enter after a heading now starts a normal paragraph instead of another heading.
• Inserting a new page or database leaves a visible block on the current page instead of jumping you away.
• Newly added blocks like columns no longer disappear after you reload the page.
• Toggle blocks now line up cleanly, and HTML blocks render properly on shared pages.
• Sidebar rows, counts, and controls align to a steady edge instead of shifting as you hover.

New

• Ask the assistant to "research my competitors" and it finds them itself, then pulls live social data.
• It reads real follower counts, engagement, and recent posts across Instagram, TikTok, YouTube, X, and LinkedIn.
• The assistant saves a competitor snapshot and compares rivals to point out your whitespace.
• Competitor data lands in your knowledge base, so it stays searchable and citable in future answers.
• Add comments to any page with a new Notion-style comment action on the page header.
• The assistant chat launcher is now available on every workspace page, database, and report.

Fixed

• The assistant now actually reads a report or file you attach, instead of replying "Evidence missing."
• Attach pages, databases, files, or reports from the side chat too - attached items stay visible after you send.
• Ask the assistant to expand a selected page and it edits that page (with approval) instead of creating a new one.
• Selecting text before opening chat now drops it into the composer as context.
• Approving an assistant page edit is more reliable and no longer saves an empty message.
• Assistant-generated snapshots and artifacts now save reliably - a silent save error dropped them before.
• Slash commands for images and databases insert cleanly without leaving stray "/image" or "/database" text.

New

• Share to web: publish any page, database, or card as a read-only link - public, or limited to your organization - and revoke it anytime.
• A new Settings → Shared links page lists every active link so you can revoke them in one click.
• Paste long or messy notes into chat and the assistant structures them into content drafts, calendar entries, and task-board cards - all behind a single approval.

Fixed

• Approving an assistant edit to a page now applies right away without a reload, and no longer fails with a false "page changed since preview" error.
• Clicking Approve twice no longer applies the same change twice.
• Fixed a workspace chat rendering glitch and tightened the sidebar and chat layout at common window widths.

New

• Every page, database, chat, and report has its own URL - reload or bookmark and you're right back in it.
• Right-click any tab to copy its link.
• Share to web: publish a page or database as a public read-only link, revocable anytime.

Improved

• The editor slash menu is now fully translated in Russian, and Ukrainian coverage is complete.
• Numbers, dates, and IDs render in the proper monospace font across the app.

Fixed

• Images inserted into pages now actually display - they never loaded before.
• Pressing Enter on a line starting with "/" now creates a new line instead of doing nothing.
• Fixed a crash for Ukrainian-language users during first-time setup.
• A failed content generation now reliably refunds your credit.
• Asking the agent to rename or delete a file now requests your approval and does it.
• The workspace logo is visible again on the dark sidebar.

New

• Drag cards to reorder them inside a column - an insertion line shows exactly where they'll land.
• Images and PDFs attached to a card now open in a full-screen in-app viewer with arrows, keyboard navigation, and a real download button.
• You can now edit your own card comments in place - edited notes get an 'edited' marker.

Improved

• Board columns now stretch to full height and scroll individually, so drop zones are always reachable.
• Long card and page titles wrap cleanly instead of spilling over the edge.

Fixed

• Fixed a stuck loading screen after a hard refresh where only the left sidebar appeared.

Fixed

• Opening an assistant-created page or task board from chat no longer crashes your workspace.

New

• After setup, the assistant speaks first - grounded in your business, never a blank chat.
• Tell the assistant about your business in plain words; it saves your profile field by field until complete.
• Up to three suggested next actions sit above the message box after every reply.
• The landing page offers two equal starts: analyze your website, or begin without one.
• If the assistant is slow to start, you instantly get a useful fallback with your profile and next steps.

Improved

• Answers to your questions are now grounded in your saved business profile.
• A completed report that already belongs to a project now offers 'Open project' directly.
• Renaming a page or conversation updates its open tab everywhere instantly.
• Arrows in the content editor now move between your platform drafts.

Fixed

• Asking 'how does this work?' no longer locks the whole conversation into one mode.
• Chat replies no longer repeat everything the assistant said earlier in the thread.
• Fixed an editor crash when working with pages and databases.
• Your first report no longer hits a 'project limit' wall when you convert it.
• The setup form no longer covers your report results after signing up from one.
• The welcome tour no longer covers the setup screen, and the chat launcher no longer blocks Send.
• Your own edits to the profile page always survive assistant updates.
• Closing the welcome tour can no longer mark setup complete before your profile exists.
• Fixed a sign-up failure when two people share the same first name.

Improved

• Signing in now takes you straight to Workspace OS - the primary home for your account.

Fixed

• New signups are guided through a first-run company-profile setup instead of a blank screen.
• The notifications dropdown no longer opens off-screen under the sidebar edge.

New

• Running low on credits? Buy more instantly with pay-as-you-go top-up packs, right where you hit the limit.
• Upgrade to a paid plan from a real in-app pricing and billing page.
• The whole app now speaks Ukrainian - pick it from the language switcher; emails are translated too.
• Dates now show in your own language across reports, content, and settings.
• New workspaces get a guided first-run that captures your company profile in one step.
• Find and index knowledge sources for a workspace directly from the OS.
• Attach your own image to a content draft instead of only generating one.
• Copy a shareable link to any database card and reopen it straight from the link.

Improved

• Strategy now renders as clean, formatted previews instead of raw data.
• 'New conversation' sits at the top of your chat list, and the Home entry is clearly labeled.
• Finished drafts get a one-click 'Add to calendar', and calendar-created drafts keep their slot.
• Removed duplicate, dead-end Strategy and Reports screens so there's one place for each.

Fixed

• The assistant now always talks about the workspace you're actually in after switching.
• A brand-new workspace asks about your business instead of inventing details.
• Credit charges and refunds now always hit the wallet you're actually using.
• The onboarding tour no longer reappears on a second device.
• Reports now show their own workspace's name instead of the active one.

New

• Open a finished report and the assistant picks up from its top findings - your report now leads straight into the workspace.
• No website? The landing page offers a second path - chat with the assistant directly instead of entering a URL.
• Start a new analysis from inside your workspace with a “Run new analysis” action in the sidebar.

Improved

• The workspace sidebar is now grouped into Work, Inputs, and Outputs, so reports, knowledge, and content are easier to find.
• The public report page now lets you sign in and continue right where you left off in your workspace.

New

• Ask the assistant to do several things in one message - it plans the steps and carries them out in order.
• The assistant now saves the content it generates as drafts you can open and edit, instead of just showing them in chat.
• It remembers what you created and brainstormed earlier, so new work builds on it instead of starting from scratch.
• The whole app is now available in Russian - pick it from the language switcher; emails are translated too.
• Relation fields can link to a specific database, picked Notion-style right when you create the column.
• Create database columns from a proper picker with a searchable, grouped type list - no more text-only prompt.
• Drag to reorder columns in the database Table view.
• As you type in chat, a matching skill is suggested above the box - apply it with one click.

Improved

• Messages sent through a skill now show a clear “skill applied” chip.
• The skills picker gained search, most-used sorting, scope badges, and an enabled count.
• The database Table view shows clean vertical column dividers, and reordering a column feels instant.

Fixed

• Switching organizations and coming back now reopens the exact workspace and page you were in.
• The database and the rest of the workspace now follow your chosen language everywhere.
• Fixed a crash when opening a draft the assistant created for you.
• Chat no longer shows raw internal data - just the assistant’s reply.
• The empty kanban “No status” column now hides itself, and several column-drag, search, and relation glitches are fixed.

New

• Ask the assistant to build task boards, pages, and databases right in chat - it creates them for you and hands back a card you can open.
• The assistant now learns from your corrections and improves its own skills over time - safely, reversibly, and never touching skills you wrote yourself.
• Assign tasks to teammates and get notified when a task is assigned to you or its due date passes.
• Databases gained new field types: URL, Email, Phone, ID, File & media, and automatic Created / Last-edited time & by.
• New here? A getting-started guide and a warmer first run help you get a real result in your first couple of minutes.

Improved

• The assistant now knows every skill your workspace has, so it suggests the right one instead of overlooking it.
• It recalls relevant context from earlier - even older notes - not just your most recent messages.
• Dates now follow your chosen language instead of always showing in one locale.

Fixed

• Fixed a production error that could blank out generated-content previews.
• Tightened access so a workspace's skills can only be seen by people in that workspace.

New

• Deleting a project now shows an Undo toast, so you can restore it within a few seconds if it was a mistake.

Improved

• Deleted projects disappear instantly from the switcher, project list, and usage count.
• You can no longer delete your last remaining project - the Delete button is disabled with a tooltip.

Fixed

• Fixed an issue where deleting the project you were viewing left you on a 'not found' page - you're now redirected cleanly.

New

• Workspace OS is now the home screen - open a workspace and you land straight in it, no analysis step required.
• Edit kanban board columns directly: add, rename, and delete sections right from the board view.
• The workspace switcher now shows how many workspaces you've used, with a one-click upgrade when you reach the limit.

Improved

• Switching organizations and workspaces is smoother - no more white flashes or a false 'No projects yet'.
• Cleaner sidebar: top-level sections collapse into an accordion, and Trash is always pinned within reach.
• Organization invites and management moved into the org rail; the sidebar header now switches workspaces.

Fixed

• Fixed crashes that could blank a pane when switching workspace or organization.
• Tightened workspace isolation so settings and data can't be reached across tenants, and access ends when you leave an org.
• Chats now name themselves from your first message and never overwrite a name you set.
• Limit messages now show the proper upgrade prompt in production instead of a generic error.
• Deleting a large workspace no longer fails on big knowledge bases.
• The page editor scrolls with your mouse wheel again.

New

• New Workspace OS at /ws/[id]/os: project navigator on the left, tabbed center pane that holds chat threads and file previews side-by-side, file system on the right. Resize the panes; layout syncs across devices.
• Upload PDFs and DOCX files directly into your workspace - drag onto the file pane or click Upload. We parse, chunk, and index them so the agent can cite them in answers.
• Chat answers now include clickable citation pills. Click a pill and the source document opens in a preview tab right next to the conversation.
• New Analytics dashboard at /ws/[id]/analytics: agent activity, chat volume, content output by platform, and knowledge index health, with a 24h/7d/30d/90d window picker.
• Reusable voice profiles for content: create per-channel style instructions (e.g. 'Arthur's Telegram voice') with exemplars, mark one default per channel, and pick from the dropdown when generating.
• Reports the agent writes for you save back to your workspace and appear in a Generated group in the file pane - click to read or share.

Improved

• New sidebar entries for Workspace OS and Analytics (both flagged beta).
• Resize handles double-click to reset; per-workspace panel widths and open tabs remembered across devices.

Fixed

• Dashboard link is now visible in the top navigation bar.

New

• You can now browse solutions by role on the Roles page.
• New Use Cases page shows growth workflows by objective.
• Blog page with filtering by format and topic.

Improved

• Redesigned floating navigation bar with improved mobile menu.
• Footer now links directly to Roles, Use Cases, and Blog.

Improved

• Sidebar now links to Changelog, Blog, About, and FAQ pages.

Fixed

• You can now re-import reports or import them to different workspaces.

Fixed

• Reports now import correctly into workspaces with all content.
• Re-importing a report no longer creates duplicate content.
• Replace mode properly clears old content before importing.
• Business name is now detected reliably during import.
• Knowledge base sources are no longer missed during import.

Improved

• You can now create new analyses without leaving the reports page.

Fixed

• Workspace switching is now faster with less screen flickering.
• Sidebar workspace selector no longer briefly goes blank.

Fixed

• Invalid report links now show a proper error instead of crashing.
• Workspace navigation no longer gets stuck in redirect loops.
• Report limit now only counts active analyses, not completed ones.

Fixed

• "Add to Workspace" now works correctly for all report types.
• Importing to personal workspace properly transfers ownership.
• Dashboard health scores now display correctly for all report formats.
• Dashboard shows the latest completed report instead of pending ones.

New

• You can now choose between creating a new workspace or replacing an existing one.
• Anonymous users can view their report before signing in.

Fixed

• Starting an analysis now shows errors instead of failing silently.
• Workspace import progress steps are now tracked accurately.

Fixed

• Workspace edits now require proper ownership verification.
• Credits are deducted before generation and refunded on failure.
• Long-running analyses now time out gracefully instead of hanging.
• Backend callbacks retry automatically on temporary failures.

Improved

• Importing a report creates a fully populated new workspace.

Fixed

• "Add to Workspace" now shows correctly for reports not yet imported.
• Failed imports reset the report so you can retry.
• Workspace import validates reports more reliably.

Fixed

• "View in Workspace" now appears when report is already imported.
• Strategy page shows the correct business name instead of "Untitled Report".
• Workspace conflict dialog shows real names and content counts.
• "View in Workspace" navigates to the correct workspace.

Fixed

• Strategy page title and buttons are no longer hidden behind the nav bar.

Fixed

• Social media audit scores now appear in your report reliably.
• SEO and social scores no longer disappear due to data formatting issues.

Fixed

• SEO audit retries automatically when initial analysis is incomplete.
• Social media audit retries when results come back empty.

Fixed

• Platform content now generates much faster using an updated AI model.
• Source research and competitor analysis retry on empty results.
• Image generation is more reliable across all providers.

Improved

• Content generation is now 10x faster with a new AI model.
• Analysis pipeline runs 50% faster with parallel processing.

Fixed

• Business discovery retries when initial detection is incomplete.
• Source research and competitor analysis pre-fetch data for better results.

New

• You can now see real-time progress when importing a report to a workspace.
• Workspace conflict dialog lets you choose to create new or merge.

Fixed

• Knowledge base sources now import correctly from reports.
• Strategy data no longer gets mixed between workspaces.
• Workspace names are now set correctly from business profile.

Fixed

• Knowledge base sources now import from all report sections.
• Customer personas now import with all profile details.
• Reports without a strategy now generate a basic summary automatically.

Fixed

• Instagram reels and TikTok scripts now display correctly.
• Blog post content renders properly in all views.
• Workspace content imports with correct image mappings.

New

• "Add to Workspace" now imports your full strategy and knowledge base.
• New "+ New Report" button on the reports page.

Fixed

• Sources section now displays all discovered categories and data points.

Fixed

• Team members can now access shared reports without authorization errors.

New

• Report limits now apply based on your plan (Free: 1, Pro: 2, Agency: 5).
• Organizations default to Agency plan with 5 workspaces and 5 reports.

Fixed

• Workspace navigation now goes to the correct workspace after creation.

Fixed

• New workspaces now correctly import business name and industry from reports.

Fixed

• Overall score is now calculated from SEO, AEO, social, and local scores.

New

• Landing page now has navigation, demo preview, pricing, and FAQ sections.
• Welcome modal greets first-time users after signup.
• You can now cancel an audit while it is running.
• Detailed progress indicators show what the analysis is doing.
• Knowledge Base has a "Find Sources with AI" button.
• You can add competitors manually in Strategy settings.
• Content supports drag and drop organization.
• Prominent "Add to Workspace" button on reports.

Fixed

• Report scores now display values instead of dashes.
• Content edit button no longer leads to "Content not Found".
• Rich text formatting (headings, bullets, quotes) now renders in the editor.
• AI content generation from reports and topics works correctly.
• Calendar cards are clickable and show color-coded status badges.
• Strategy sections have better spacing and no longer clip content.
• Dashboard no longer shows duplicate workspaces.
• Landing page stats show real numbers.

New

• You can now generate background videos for TikTok content with AI.
• Video generation status shows on content cards (generating/ready).
• Global notifications appear when your video is ready.

Fixed

• Image upload errors now show a clear message instead of failing silently.

Fixed

• TikTok content now displays correctly instead of showing a loading state.

Improved

• Target audience pain points expand fully without scrolling.
• Action plan table columns no longer overlap on narrow screens.
• Action plan rows show editable details for richer context.
• KPI tools align cleanly in a grid layout.

New

• Content library now uses a card grid layout (2-3 columns).
• Larger "New Content" button with a plus icon.
• Calendar cards have an edit icon linking to the content editor.
• Action plan rows expand to show step-by-step details.
• Reports page shows upgrade prompt when limit is reached.
• Knowledge Base stats now include helpful descriptions.

Fixed

• Content edit no longer shows 404 for workspace content.
• Headings, bullets, and quotes now render in the rich text editor.
• AI generation no longer requires a topic when using report data.
• Report scores show "Analyzing..." while in progress.
• Calendar cards have consistent sizing and status badge colors.
• Strategy nav no longer hides content behind the sticky header.
• Action plan columns no longer overlap.

New

• You can now download your strategy as a PDF or JSON file.
• "Add to Workspace" button added to the report summary page.

Fixed

• Workspace profile auto-fills from your report on completion.
• "Create Workspace" button hidden when report already has a workspace.
• Landing page links now navigate to the correct pages.

New

• Each workspace now has its own URL for easier sharing and bookmarking.
• "Create Workspace" button added directly on report pages.

Improved

• Workspace selector navigates to the correct scoped URL on switch.

Fixed

• New reports no longer overwrite your existing workspace strategy.

New

• Strategy edits are saved to your workspace and persist independently.
• First edit auto-copies report data so you can customize freely.
• Regenerating a strategy from a report syncs updates to your workspace.

Improved

• Reports are now immutable snapshots that your edits cannot overwrite.

Fixed

• Calendar entries now have proper ownership checks before changes.

Improved

• Brand Voice is now a prominent full-width section with rich text.
• Content Pillars and Platform Priorities display side by side.
• Pain points use a multi-line field for better readability.
• Hover states added to persona cards, budget tiers, and action plan.
• Budget "Recommended" tier is now detected from your data.

Fixed

• Action plan hover states are now visible.
• Platform priority buttons have larger click targets.
• Roadmap timeline dots display as circles instead of squares.

New

• Workspace target audience and brand voice auto-fill from your report.
• Strategy page now has smooth animations and section navigation.

Fixed

• Regenerating a strategy no longer changes your report status.

New

• Strategy now includes brand voice, action plan, budget tiers, and roadmap.
• Competitor positioning matrix with tactics and differentiators.
• KPI framework with target tables and recommended tools.
• "Regenerate Strategy" button to refresh your strategy from the latest data.
• You can now invite team members and manage invitations from the sidebar.

Fixed

• Platform priorities and content pillars display correctly.
• Target audience and brand voice sync properly to new workspaces.

Improved

• Account and organization menus now match the sidebar design.
• "Manage Account" and "Manage Organization" open in clean modal dialogs.
• Notification bell is now easier to click.

New

• You can now create a workspace directly from a completed report.
• Workspace logos appear in the workspace selector.

Improved

• Workspace selector moved to sidebar footer for easy access.
• Business info and workspace profile merged into one settings card.
• Dashboard shows an editable workspace card instead of read-only summary.

Fixed

• Location field no longer shows raw code on the settings page.

Improved

• Scores are now larger with color-coded borders and animations.
• Score bars show color-coded fills that animate on load.
• Sticky navigation has larger, bolder active state indicators.
• Typography uses a display font for section headers.
• Hover states added to competitor, local business, and content cards.
• Copy buttons appear on hover for a cleaner reading experience.

Improved

• Report width increased for better readability.

Fixed

• Sticky nav now tracks your scroll position correctly.
• Sticky nav is no longer hidden behind the header on report pages.
• Empty sections no longer appear in reports.
• LinkedIn content renders correctly for all data formats.
• Duplicate content blocks no longer appear in reports.

New

• Team changes (member added/removed, org updates) now sync in real time.
• Deleting a team removes all team workspaces and reassigns members.

New

• Content previews now match each platform (Twitter, Instagram, LinkedIn, TikTok, GBP).
• Reports are now viewable inside the dashboard without leaving the app.
• Get an email notification when your report is ready.
• Email notification prompt appears during report generation.

Fixed

• Blog post content and social sections now display correctly.

Fixed

• Reports no longer leak between personal and team contexts.
• Analysis data is now properly scoped to the correct workspace.

Improved

• Workspace and content edits now verify ownership before saving.
• Strategy editing requires authentication.

Fixed

• Analysis pipeline now recovers gracefully from individual agent failures.
• Failed analyses are now reported to the user instead of hanging forever.

New

• Landing page now includes social proof stats with animated counters.
• Problem-solution section highlights key pain points.
• Step-by-step transformation timeline (Analyze, Strategize, Create).
• Features grid showcasing platform integrations.
• Free beta pricing card with feature checklist.
• FAQ section with 8 common questions.

Improved

• All sections fade in on scroll with motion-safe animations.
• Fully responsive layout for mobile and desktop.

Improved

• Cross-platform strategy receives richer content summaries.
• Visual content now considers competitor themes for differentiation.

Fixed

• Competitor analysis now includes content gaps and opportunities.
• Overall score now factors in local presence and competitor data.
• Parallel analysis failures are caught and reported clearly.

New

• Reports load progressively with section-by-section skeleton placeholders.
• Sticky nav shows loading indicators for sections still being analyzed.
• "Sign in to save" banner for anonymous users viewing reports.
• Reports auto-save to your account when you sign in.
• "Copy to Team" button lets you share reports with your organization.
• Platform content tabs for Twitter, Instagram, LinkedIn, TikTok, and GBP.

Improved

• All report URLs now use a consistent /report/ path.

Fixed

• Nine content fields now save correctly to your report.
• Calendar data no longer leaks between personal and team contexts.

New

• You can now assign existing content to calendar time slots.
• "Create content" button in calendar pre-fills the date and time.

Fixed

• AI content generation now uses your workspace context for better results.

New

• Sort content by newest, oldest, or recently updated.
• AI-generated content is marked with a purple "AI" badge.

Fixed

• AI-generated social posts now appear in your content library.
• AI generation no longer creates duplicate blog posts.
• Content editor resets properly when switching between items.
• Error and success messages now appear during content operations.

Improved

• New serif + sans-serif typography pairing for a refined editorial look.
• Warm stone color palette with terracotta accent throughout the app.
• Dashboard redesigned with the new visual aesthetic.
• Buttons, cards, and badges updated with sharp corners and new palette.
• Sidebar navigation uses left border indicators for active state.

New

• You can now set up an organization profile with industry and brand voice.
• AI content generation uses your profile for more relevant output.
• Inline correction UI for business profile fields in report view.

New

• Your first name or email now displays in the sidebar footer.

New

• App version is now visible in the sidebar footer.

Fixed

• Content page no longer flickers when loading for the first time.