What’s new
What has shipped in TakeOne, newest first, from version 1.0 on 4 September 2026. Follow it in a feed reader with the Atom feed.
- Each app you connect through the MCP server, such as Claude Code, Cursor or Codex, is listed by its own name under Connected apps, and a take’s history says which app changed it.
- Stop one connected app on its own and allow it again later, while your other apps keep working.
- Developers opens with Connect an app: pick your app, copy one line, and connect your account once for every app.
- Connected apps stay signed in when TakeOne’s MCP server is updated, instead of asking claude.ai to sign in again.
- Your public page is part of Starter and Pro. On Free it is hidden, not deleted, and comes back as it was when you upgrade.
- On a wide screen, a phone-shaped take no longer pushes the Share panel down the recording page.
- Connected apps show each tool’s own logo, and Connect an app picks from four: Claude, Codex, Cursor or another app.
- A revoked token leaves the tokens page after a few seconds instead of staying in a Revoked list.
- The dot beside What’s new also lights when something is added to a day you already read.
- A trimmed or sped-up take’s share page, link previews and public page show the length it plays, and the take.published webhook carries it too.
- Choose a take’s poster frame in the editor or through the API; publishing makes the new poster without rendering the video again.
- An agent can ask you for a take with one link: it opens the recorder with the title and script filled in, and files the take where the agent asked.
- Agents can read a take’s view count, its render’s progress, and the timeline strip of 24 frames across the take.
- Agents can look at a take: frames at the moments they choose, up to 12 at a time, so they can see what is on screen before they cut or zoom.
- A failed render now sends the take.failed webhook, saying it was the render that failed while the take still plays.
- A recording request read through the API now says who asked for it, as the recorder already showed.
- Every MCP tool has a readable title, and the ones that overwrite, delete or publish are marked so your agent’s app asks before using them.
- Pages anyone can read about TakeOne: its features, how agents work with a take, security, questions, pages for founders, engineers and developer advocates, and this changelog.
- What’s new at the foot of the sidebar lists what changed, with a dot when something arrived since you last looked.
- TakeOne for Mac is available to download from takeone.dev: record a display, a window or an app from the menu bar, with your microphone, the system’s sound, your camera and your clicks.
- The Mac app has a library window and an editor of its own: trim, cut, zoom, frame and reshape a take, place the camera bubble, draw the cursor, speed up idle stretches and publish.
- A new Get the apps page lists the places to record from, the Chrome extension and the Mac app beside the browser, linked from the front page and the app.
- Sign in with Apple, beside Google, on the sign-in page.
- Auto-zoom now meets back-to-back zooms at the right point between their clicks.
- Finding idle stretches no longer misses still stretches in takes recorded with few frames.
- Through the API, an agent can search your takes by title and page through the library, with each take’s link and poster in the list.
- The TakeOne extension is on the Chrome Web Store: record the tab you’re on, clicks included, on any site and without adding a script.
- Add to Chrome appears on the record page, the front page, onboarding, Help and the support page, and the record page shows when the extension is installed.
- Recording quality follows the plan: 1080p at 30 or 60 fps on every plan, with Free’s 60 fps takes stopping at 5 minutes, and 1440p on Starter and Pro.
- Publishing says how long the render will take, in the editor and through the API, and 1440p takes at 60 fps are capped at 20 minutes with the camera and 23 without, so every render finishes.
- The editor’s timeline shows its thumbnails at once, even on long high-resolution takes.
- Sign-up is open: anyone with a Google account can start on takeone.dev.
- On a phone, the record page says recording needs a browser on a computer instead of failing at Start.
- A public support page says how to get help.
- Takes recorded in Chrome no longer open frozen on the countdown or start late.
- The account menu sits at the foot of the sidebar, and the sidebar’s toggle at the top left.
- Codex joins the clients you can connect to TakeOne’s MCP server, with its setup command on the Developers page and in onboarding.
- A private take is now visible to its owner alone, and the setting reads Only me.
- A new Account and settings page gathers your account, your public page, your plan and your API tokens, and pages use the width of a wide screen.
- Agents connected by sign-in stay connected after their access refreshes.
- A share link can carry a start time, opening the take at that moment.
- A link to the recorder can name the take in advance, and link previews carry the take’s running time.
- Through the API, an agent can find the take behind a share link.
- The MCP server’s tools are named exactly as the API’s operations.
- Delete your account and everything in it from Account and settings.
- A cancelled plan shows on the billing page as ending on its date, instead of renewing.
- Before you add the beacon, TakeOne reads your page and names the install that fits it and any content security policy that would block it.
- The beacon can run only on the hosts you name, so one tag can sit in a production layout, and bunx @takeone/beacon install adds it for you.
- A pinned copy of the beacon and a data sheet are ready for a security review.
- Renders that run without your tab open rise to 60 minutes a month on Starter and 300 on Pro.
- A first sign-in starts at a short welcome that asks how you’ll use TakeOne and walks through three steps that fit the answer.
- A Getting started list stays in the library, or as a pill beside the account menu once hidden.
- Move every take in a folder to another in one step.
- Unlisted share links are kept out of search engines.
- Plans: Free, Starter and Pro, monthly or yearly, with a billing page that shows your plan and usage and opens the billing portal.
- Each plan sets its limits on storage, take length, protected links, publishing from an agent, agent calls a day, webhooks and render minutes, with a banner as you near one.
- The library gets search, folders, sorting and filters, a list view, drag and drop, and a selection bar to move or delete several takes at once.
- Agents can list folders and file takes into them through the API and the MCP server.
- A new front page shows a take recorded in Chrome and finished by your own agent, with the plans and a yearly price switch.
- Connect an agent by signing in to TakeOne and approving it, instead of pasting a token, and disconnect it from the tokens page.
- The agent connect page opens already signed in from TakeOne, with no second account to make.
- A Developers section in the sidebar gathers the MCP server, the API, webhooks, the beacon, tips for agents and examples.
- Send feedback from the sidebar, and read the terms and privacy pages, with plans, payment and refunds set out.
- Webhooks tell an agent or your own service when a take is ready, published or failed, as signed requests set up from the tokens page or the API.
- An agent can publish a take through the API: your open TakeOne tab renders it, or TakeOne does when no tab is open.
- A floating window keeps the timer, your camera and Stop above whatever you’re recording, in Chrome and Edge.
- Pause and resume a take, with P or from the floating window, and the paused stretch is left out of the file.
- In Safari, the camera floats in picture-in-picture and the media keys pause and resume.
- Takes published from Safari keep their zooms, their cursor and a round camera bubble.
- Details and the recorder say which encoder records a take, and what the quality setting can and can’t control in Safari.
- The beacon is on npm as @takeone/beacon, with a script tag, a React hook and a Vite plugin, and the recorder shows how to add it when a page has none.
- Pair your app with the recorder by a code, without opening the app from the recorder.
- Recording in Safari starts from the first click and includes the camera bubble.
- When a saved camera or microphone is gone, the recorder uses the default one and says so, instead of stopping.
- Ready-made skills for your agent are published: describe a take, draft its script, place zooms from clicks, tighten silences and transcribe.
- Record your microphone and the tab’s or system’s sound, mixed into one track, with a level meter while recording and volume on the player.
- Zoom into a take by hand or from your clicks in apps that carry the TakeOne beacon, with a drawn, smoothed cursor and click ripples.
- Show your camera as a bubble in a corner of the take, with its size, shape and mirroring in the editor.
- Reshape a take for a feed or a story, speed up its idle stretches, and save a GIF from the playhead.
- Give each take a public description and a private script; a transcript adds captions and chapters to the share page, and large uploads resume after a reload.
- Your own agent can work on your takes through API tokens, the public API and the MCP server, and each take shows what an agent last changed before you publish.
- TakeOne 1.0: record your screen, a window or a tab in the browser, up to 30 minutes, and get a link to share.
- A take is kept on your computer while it records and uploads, so a closed tab or a crash can be recovered.
- Trim, cut and frame a take in the editor, then publish the edited version or download it.
- Share links can be kept private, protected by a password or set to expire, and unfurl in Slack and iMessage with a title, a poster and a short preview clip.
- Viewers watch in any modern browser with seeking, speed and full screen, and each take counts its views.
For how things work now rather than when they arrived, read what TakeOne does.