How LitmusRun works

LitmusRun is an AI agent product for synthetic user validation and QA. A panel of agents opens your web app in an isolated Chromium browser, acts like your target users, and returns ranked findings with session evidence.

The loop

1. Paste your URL, set the mission, and choose how deep the study should go.
2. AI agents browse, click, hesitate, and rage-tap in real time as synthetic users.
3. Live thoughts and session frames stream to the agent room.
4. The report lists validation and QA findings, persona takes, and a consensus summary of what to fix first.

What you can test

• Web apps and user flows on staging or production, with heatmaps, session evidence, and live thoughts.
• Full products, single features, or paths like signup and checkout.

Get started

Sign up free with 20 credits, paste a product URL, set the mission, and launch a panel study from the dashboard. Upgrade anytime for more monthly credits and larger agent panels.

Start for free →

CI & API

Launch studies from CI or scripts with a profile API key. Create keys in Profile → API.

curl -X POST https://litmusrun.com/api/v1/studies \
  -H "Authorization: Bearer lr_<your_api_key>" \
  -H "Idempotency-Key: deploy-$UNIQUE_ID" \
  -H "Content-Type: application/json" \
  -d '{
    "projectId": "<project_id>",
    "missionGoal": "Complete signup without reading the docs",
    "studyType": "standard",
    "agentCount": 12
  }'

curl https://litmusrun.com/api/v1/studies/<study_id> \
  -H "Authorization: Bearer lr_<your_api_key>"

• Authorization — Bearer lr_… profile API key from Profile → API.
• Idempotency-Key — required unique string per launch (retries reuse the same panel run).
• Body — projectId, missionGoal; optional artifactUrl, studyType (smoke | standard | deep), agentCount (3–80).

Prefer the dashboard for interactive launches? Open the launch wizard →

REST API reference →