Start here

Record your first agent task.

Record lets AI agents capture desktop work as a video you can watch, comment on, and share.

Shortest path

Paste the install-and-login command into your agent.
Sign in or create a workspace in the browser window, then approve the named agent.
Ask the agent to record a specific task and duration.

Record ten seconds

record-cli start --duration 10 --title "Homepage smoke test" --agent codex --json

Success: the JSON result contains a recording URL and the Library shows a playable video.

One time per agent

Sign in through the browser and connect the agent.

record-cli login opens Record in your browser. After you sign in and approve access, the CLI saves a named, revocable agent identity on that machine.

**Approve scoped agent access.**Your account password stays in the browser and administrator credentials never enter the prompt.Open GIF

Review before you approve.

The browser shows the requested agent name, workspace, allowed actions, and expiry. Revoke that identity from Agent access at any time.

For each task

Ask for a concrete recording.

Name what should be demonstrated and how long to capture. The agent records the desktop and publishes the finished video.

**Record and publish from the CLI.**The structured result confirms capture and includes the finished recording URL.Open GIF

Example request

“Record a 10-second mobile smoke test of the homepage, publish it to Record, and return the recording link.”

Product objects

Live run: An active Pro recording stream that becomes a recording when it finishes.
Recording: The finished video, timeline, attachments, transcript, and comments stored in the Library.
Agent access: A named, scoped, revocable credential for one agent or service.
Recording link: A revocable viewer URL limited to one recording.

Sign in through the browser and connect the agent.

Ask for a concrete recording.

Review and share the recording.

Product objects