Watch the agent run live — and replay the recording later
Stream the browser session in real time, then keep a durable video of every run for debugging, audit, and proof of what happened.
GET /api/v1/runs/{id}/liveWhat live view & session video does
Headless automation is a black box until it breaks. Twin streams a live view of the browser as the agent works, and records each session to durable storage so you can scrub back through exactly what the agent saw and did. Observability is built in, not bolted on.
Real-time live view
Watch the cursor move, pages load, and steps execute as they happen — from the dashboard or the API.
Durable session video
Every run is recorded and stored durably, so you can replay it long after the run finished.
Tied to the step list
The video is aligned with the run's step list, so a failure jumps you straight to the frame where it happened.
Proof for audits
A recording plus machine-readable headers is concrete evidence of what an authorized run actually did.
From a goal to deterministic action
- 1Start a runKick off a goal or a skill; Twin opens a real browser session for it.
- 2Stream liveSubscribe to the live view to watch the session render in real time.
- 3Record to storageThe session is captured and written to durable storage as it runs.
- 4Replay anytimeFetch the saved video and step list later for debugging, demos, or audit.
See it on a real call
Subscribe to a live stream while the agent runs, then fetch the durable recording and step list afterward.
const run = await twin.agents.run({ goal, url });
// Watch it happen
const live = twin.runs.live(run.id); // real-time stream
// …or replay it later
const { videoUrl, steps } = await twin.runs.get(run.id);
// → durable mp4 + aligned step list- Start a rundone
- Stream liverunning
- Record to storagequeued
- Replay anytimequeued
What live view & session video is
The facts — how it works, what it costs, and the signal you get back on every call.
| Property | Twin Browser |
|---|---|
| Live view | Real-time session stream |
| Recording | Per-run session video |
| Storage | Durable object storage |
| Alignment | Synced to step list |
| Access | Dashboard + API |
| Use | Debug, demo, audit |
Live view & session video — common questions
Can I watch a run while it happens?
How long are recordings kept?
How does video help debugging?
The rest of the platform
Deterministic replay
A successful run compiles into a skill — an ordered, parameterized program of browser actions that replays the same way every time.
Human-in-the-loop handoff
Approvals, MFA on an authorized flow, or an ambiguous choice pause the run, hand off to a person, and continue from the exact same point.
Agent & skill library
Every compiled skill is stored, versioned, and reusable — browse them, run them directly, and share them across your agents.
Make every run cheaper than the last.
Start free, compile your first skill, and watch the marginal cost per run trend toward zero as your agents reuse what they have already learned.