pyagentbrowser 0.31.2rc0

Python SDK for controlling the native agent-browser engine in-process

pyagentbrowser

pyagentbrowser embeds the native Rust agent-browser engine in Python. It gives agents accessibility snapshots, stable element refs, and before-and-after evidence for browser actions.

uv add pyagentbrowser

from agentbrowser import Browser

browser = Browser()
browser.open("https://example.com")
page = browser.observe()

print(page.text)

Browser starts lazily and stays active until browser.close() is called. Keep the same object across notebook cells or interactive Python commands. browser.session.status() reports browser and restore lifecycle state. close() returns the terminal save result. Screenshots returned by browser.capture.screenshot() render inline in notebook frontends.

Snapshot binds accessibility refs to the page state that produced them. Ref actions return an ActionResult containing the original snapshot, the resulting snapshot, and their diff.

Use browser.find for live CSS, XPath, and semantic queries. Namespaces such as browser.capture, browser.tabs, browser.network, and browser.cdp expose focused browser capabilities. Every action in the pinned native engine remains available through browser.native.

Documentation

• Get started
• Guides
• API reference
• Runnable examples

Installation options

The distribution is named pyagentbrowser. Python imports it as agentbrowser. Wheels support Python 3.11 through 3.14 on macOS, Linux, and x86-64 Windows.

uv add "pyagentbrowser[images]"  # Pillow-backed screenshot helpers
uv add "pyagentbrowser[cdp]"     # CDP frames, targets, and evaluation

Local launches find an installed Chromium browser or prepare Chrome for Testing through the native installer.

Development

make install
make check

make test-integration exercises the Python, PyO3, adapter, and CDP seams against a real browser. make check-release builds and installs both distribution artifacts.

License

Apache-2.0. See LICENSE and NOTICE.