Made for parallel agents
Several worktrees, several sessions, one of you. Finished tasks queue up faster than you can click through them — recordings let you verify asynchronously instead of dropping everything to re-run a flow by hand.
AGENT-RECORDED PROOF
When an AI agent finishes a task, it records the browser run, uploads it over MCP, and hands back a link. No more clicking through your agents’ work — watch the proof at your own pace while the other sessions keep moving.
Every recording is an agent driving a real browser — cursor, clicks and captions. This one is an agent walking through this very app.
Several worktrees, several sessions, one of you. Finished tasks queue up faster than you can click through them — recordings let you verify asynchronously instead of dropping everything to re-run a flow by hand.
Your agent finishes while you are away — on the long road to a vacation spot, in a waiting room, out with the kids. The link opens on a phone: watch the run, approve it, or send it back to rework. The rest stays rest, and the work keeps moving.
Upload is a tool call at the end of the agent’s cycle, using whatever browser automation it already runs. No SDK to integrate, no CI plumbing to maintain.
Every recording expires 14 to 60 days after upload, depending on the project’s plan. Your verification stream stays small and current instead of growing into a video graveyard you pay for.
Projects work the way they do on GitHub: invite the team as owners, members, or viewers, and keep recordings and API keys grouped per project.
Anyone holding a viewer link can watch it, so never record credentials or anything else you would not publish. When a recording expires, the link dies with it.
The free plan holds 25 active recordings and 1.5 GB, keeps each one for 14 days, and asks for no card.
Sign in, create a key, and your agent’s next task ends with a video instead of a promise.
Start free