Headless browser for agents.
Open pages, click buttons, fill forms, capture DOM and screenshots — without ever opening a tab. x402-native browsing sessions.
# Start a session $ curl -X POST https://browse.prim.sh/v1/sessions \ -H "X-402-Payment: $TOKEN" \ -d '{"url": "https://example.com"}' { "id": "sess_k7x9m2", "status": "ready" } # Run a script $ curl -X POST https://browse.prim.sh/v1/sessions/sess_k7x9m2/eval \ -H "X-402-Payment: $TOKEN" \ -d '{"script": "document.title"}'
Playwright backend
Headless sessions
DOM + screenshots
Part of agentstack

What agents use it for

Scraping

Render JavaScript-heavy sites with full browser semantics. Extract DOM, text, and screenshots in one call.

Form flows

Log in, click through onboarding, and submit forms. Agents manipulate the browser as if they had hands.

Visual QA

Compare screenshots over time. Catch layout regressions before humans notice.

Audit trails

Store screenshots and DOM snapshots as evidence of what the agent actually saw.

API reference

POST   /v1/sessions                 # Start session
POST   /v1/sessions/:id/eval       # Run script
GET    /v1/sessions/:id/screenshot # Capture screenshot
DELETE /v1/sessions/:id            # Close session

Pricing

ActionCostNotes
Session start$0.01Includes 60s runtime
Extra minute$0.01Billed per second
Screenshot$0.001Per capture

Browsing is an API call.

Agents don't click. They script.

Read the docs →