Use case
QA & test automation
Author end-to-end tests as goals, run them deterministically, and replay every failure as session video.
The problem
What the status quo costs you.
End-to-end test suites are expensive to write and flaky to run. Selector churn breaks tests on every redesign, and when a run fails in CI you get a stack trace, not a recording of what the browser actually did. Maintaining the suite often costs more than the bugs it catches.
See how Twin worksThe wedge
Compile once. Replay forever.
Twin lets you express a test as a goal, compiles it into a deterministic skill, and replays it the same way every time — so a pass is a pass and a failure is reproducible. Because skills plan over indexed DOM state rather than raw selectors, a cosmetic redesign doesn’t red-bar the suite. Every run is captured as live view plus durable session video, so a failure is a recording you can scrub, not a guess.
- 01Write each scenario as a natural-language goal; Twin compiles it into a replayable skill.
- 02Run skills deterministically in CI via the REST API with a Bearer key — no flaky LLM-on-every-step loop.
- 03A redesign shifts the indexed DOM map; intent-matched steps survive cosmetic churn instead of snapping on selectors.
- 04Every run records to durable session video and real-time live view for instant triage.
- 05Authenticated test accounts live in the credential vault, so login flows are covered without secrets in your test code.
Outcome: A flaky, selector-bound suite becomes a set of intent-driven skills that survive redesigns, with every failure reproducible from session video — illustratively cutting test-maintenance churn while keeping per-run cost low through deterministic replay.
In practice
The actual call, and what it returns.
Compile the scenario once, then replay it from CI. A replay executes a stored path — no planner, no model in the loop — so the same test produces the same steps.
# 1 — compile the scenario once (discover, then minimize the path).
curl https://twin-browser.com/api/v1/skills \
-H "Authorization: Bearer $TWIN_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "target": "https://staging.example.com",
"goal": "buy the sample product with the test card",
"as": "checkout-happy-path" }'
# 2 — replay it from any CI runner that can hold a Bearer key.
curl https://twin-browser.com/api/v1/skills/checkout-happy-path/run \
-H "Authorization: Bearer $TWIN_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "target": "https://staging.example.com", "account": "test-account" }'
# → 200 { "success": true, "steps": 7, "runId": "…", "credits_charged": 1 }
#
# the recording of that exact session: GET /api/v1/runs/{runId}/videoWhat this call does
- Compile is the expensive call (a discovery pass); every replay after it is the flat replay rate.
- A replay binds its inputs per call — params, vault secrets, and the account session — so one skill covers every test account.
- The run id in the reply is the handle for the video, so a red build links to the session that failed.
- Unresolved parameters fail loudly before anything is charged, rather than typing a literal placeholder onto a live page.
Under the hood
The machinery under every run.
Every use case runs on the same primitives: a token-efficient view of the page, deterministic replay, and a checkpoint where a person decides.
Semantic dispatch cache
Re-phrased requests fuzzy-match a skill you already compiled, so they skip the planner LLM entirely.
Read the mechanism — Semantic dispatch cacheDeterministic replay
Matched skills replay the same way every time — a pass is a pass, and the marginal cost trends toward zero.
Read the mechanism — Deterministic replayToken-efficient DOM state
A live page becomes a compact, numerically-indexed map of interactive elements instead of raw HTML.
Read the mechanism — Token-efficient DOM stateHuman-in-the-loop handoff
Blocked steps — approvals, MFA on an authorized flow — pause for a person, then resume cleanly.
Read the mechanism — Human-in-the-loop handoffFAQ
QA & test automation on Twin — common questions.
Is Twin a replacement for Playwright or Cypress?
How do I debug a failed test run?
Can tests run in CI?
Related
More ways teams use Twin.
AI agents
Give your AI agent a real browser it can drive — reach any site, act under your guardrails, and let repeated work replay at near-zero cost.
Internal workflow automation
Automate the internal tools and vendor portals that have no API — with audit logging and human approval built in.
Accessibility automation
Drive web tasks on a user’s behalf and audit pages for accessibility — over a token-efficient view of the live DOM.
QA & test automation
Put the work on autopilot.
Start free, hand your agent the work under your guardrails, and let the runs you repeat replay from a compiled skill.