Anchor Browser agent.task() blocks with no timeout; browser_task() returns a Future plus Playwright access

Pick the right Anchor agent call for the job. client.agent.task() blocks until the natural-language task finishes server-side - it sets no HTTP timeout and does no retries because tasks run for minutes, so wrap it in your own timeout if you need one. client.agent.browsertask() returns immediately with the session identifier, a Future for the result, and a Playwright connection to the same session - use it when your code needs to drive the browser in parallel with the agent (filling a form the agent cannot see, watching progress). task() for fire-and-forget, browsertask() when you want to participate.

Context: Official Anchor Browser Python SDK README (anchor-browser-sdk-python): client.agent.task('...') creates a session (or reuses one when you pass its session identifier) and runs a natural-language task in it server-side through the perform-web-task API, then waits for the result - with no HTTP timeout and no retries, since agent tasks can run for many minutes. client.agent.browser_task(...) instead starts the task and returns immediately with the session identifier, a concurrent.futures.Future for the result, and a Playwright connection to the same session, so your code can drive the browser while the agent works.

Source: Official Anchor Browser Python SDK README (anchor-browser-sdk-python): client.agent.task('...') creates a session (or reuses one when you pass its session identifier) and runs a natural-language task in it server-side through the perform-web-task API, then waits for the result - with no HTTP timeout and no retries, since agent tasks can run for many minutes. client.agent.browser_task(...) instead starts the task and returns immediately with the session identifier, a concurrent.futures.Future for the result, and a Playwright connection to the same session, so your code can drive the browser while the agent works.

Verified resolution: Pick the right Anchor agent call for the job. client.agent.task() blocks until the natural-language task finishes server-side - it sets no HTTP timeout and does no retries because tasks run for minutes, so wrap it in your own timeout if you need one. client.agent.browsertask() returns immediately with the session identifier, a Future for the result, and a Playwright connection to the same session - use it when your code needs to drive the browser in parallel with the agent (filling a form the agent cannot see, watching progress). task() for fire-and-forget, browsertask() when you want to participate.

Source URL: https://github.com/anchorbrowser/anchorbrowser-sdk-python/blob/HEAD/README.md