Continue website work with the right saved account

Browser agents with saved logins

A browser agent can reuse a saved browser profile bound to its exact service connection. Create the profile, complete login through the live browser and inspect its saved state before relying on it. Use a separate connection for another account. Sessions can expire or require human verification, so saved cookies are not a permanent-login guarantee.

Help a browser agent stay logged in

  1. Create the exact browser connection

    Select Browser Use and create a saved profile on the intended connection. Keep each account on a separate connection; do not select an account by an ambiguous alias or list position.

  2. Complete login in the live browser

    Use human takeover for login, 2FA or CAPTCHA. Never put passwords in the task prompt. Inspect profile metadata afterward; the presence of a saved cookie domain is not proof that the session is still authenticated.

  3. Run a narrow website task

    Use the same connection and public execution identity for the task and its follow-up. Review the actual result. If the site requires another login, stop and complete that interaction instead of claiming success.

  4. End the session and manage retention

    Stop the browser session after the task. Retained media and downloads have separate consent and limits. Reconcile an unknown profile creation against its retained ID instead of creating another profile.

Documented saved-profile setup excerpt - source documentation
connection = client.tools.create(mcp_tool="browser_use")
binding = client.tools.create_browser_profile(connection.id, name="My browser")
binding = client.tools.get_browser_profile(connection.id)

Initialize client as an AgentClient from the Python quickstart. Inspect the retained binding before relying on it. The source also documents reconciliation after an unknown creation outcome.

What you get

Illustrative output - example data
profile: saved profile on the selected connection
login_check: authenticated after human takeover
task: retrieve the permitted report
result: returned report reference
cleanup: browser session stopped

Where the result goes

Read the browser task result and deliver an approved report through a configured destination. Browser recordings and downloaded files are separate artifacts with explicit retention/consent behavior; do not assume every session produces a permanently stored recording.

Cost and operating limits

Browser session usage is priced separately from agent tokens and other tools. Stop sessions when the job is complete and keep tasks narrow. Check current pricing and the selected connection mode before a paid run; saved login state does not make an open browser session free.

Check current pricing

Choose the right approach

Implementation choices for this job
ApproachWhat to evaluate
FlyMy.AI Browser UseA browser tool inside an agent workflow with connection-bound profiles and human takeover for login.
A direct website APIPrefer a supported API when it provides the required operation and reliable authentication without browser interaction.
Browserbase or Steel[[FACT NEEDED: current Browserbase and Steel profile persistence, integration and pricing comparison from approved sources.]]

Frequently asked questions

Will the browser stay logged in forever?

No. Sites can expire sessions, revoke cookies or request another verification. Check the actual authenticated state before relying on a task result.

Can two customer accounts share one saved profile?

Use a separate connection for another account. The saved profile is bound to the selected connection.

Can I put a password in the prompt?

Use the live browser takeover for login and verification. Do not put passwords in prompts or published examples.

Related jobs

Sources

Sources checked 2026-10-23. Examples are documentation excerpts, not a claim that a paid run has been performed.