chrome-bridge-sdk

Python SDK for using chrome-bridge directly without an MCP client.

See the changelog for user-visible changes.

from chrome_bridge_sdk import ChromeBridge, SessionStatus

chrome = ChromeBridge(status_callback=lambda status: print(status.value))

async with chrome.session() as session:
    tabs = await session.browser_tabs()
    await session.browser_tab_select(tab_id=tabs[0].id)
    snapshot = await session.browser_snapshot()
    print(snapshot.title, snapshot.snapshot)

ChromeBridge stores configuration only. Entering session() discovers or starts the shared managed server, waits for the extension and FIFO lease, maintains its heartbeat, and releases it on every context exit. SessionStatus callbacks can drive application logs or startup UI.

The 21 high-level methods return frozen typed models such as Tab, Snapshot, Screenshot, ConsoleEntry, and Recording. Python parameters use snake_case, including browser_drag(start_element=..., start_ref=..., end_element=..., end_ref=...); browser_type(..., submit=False) does not require the common default explicitly. Screenshot.image_bytes decodes its Direct API base64 data.

For LLM adapters, session.tool_definitions() returns JSON Schema dictionaries and session.call(method, arguments) returns the raw Direct API JSON result. SDK exceptions expose code, retryable, and outcome_unknown attributes. Calls with an unknown outcome are never automatically retried.