Observe
Structured page text and revision-scoped element handles. Read what changed before deciding what to do next.
Let your agent use the browser.
Keep the machine out of reach.
An API-first browser runtime with scoped actions, structured observations, read-only live viewing, and inspectable evidence. Bring your own model and orchestration.
Structured page text and revision-scoped element handles. Read what changed before deciding what to do next.
Navigation, forms, tabs, keyboard, scoped HTTP, and controlled page evaluation through one command protocol.
Screenshot artifacts with SHA-256 integrity, explicit command outcomes, and events you can inspect.
Use the Python client, JavaScript, or HTTP. Keep your agent framework, inference provider, and resource budgets.
Agent integration guidewith Client(endpoint, api_key=key) as browser:
session = browser.create("https://example.com")
try:
result = browser.run(
session["id"], "navigate",
{"url": "https://example.com"}
)
print(result["result"]["title"])
finally:
browser.close(session["id"])Release notes appear here after a release is published on GitHub.
View releases on GitHub