maestro/docs/tools/browse-sessions.md
clade 7049a874f3 feat: initial public release (MAESTRO v0.1.0)
Open-source release of MAESTRO, an agent orchestration platform that runs
LLM-driven tasks through sandboxed tools, with a web UI. Apache-2.0.
See README.md and docs/ (getting-started, configuration, architecture).
2026-06-03 04:01:14 +00:00

2.1 KiB

Browser Sessions

Save a logged-in browser session per site so scheduled tasks can scrape authenticated pages without you being present.

How to add a session

  1. Open Settings → ツール設定 → Browser Sessions.
  2. Click Add site session.
  3. Fill in:
    • Label: human-readable name (e.g., "My Twitter").
    • Start URL: the page that proves you're logged in (e.g., https://twitter.com/home).
    • Logged-in selector (optional): a CSS selector that only exists when logged in.
    • Login URL pattern (optional): a glob that matches the site's login page (e.g., https://twitter.com/i/flow/login**).
  4. Click Open login window — a browser appears inside the dialog.
  5. Log in normally. Solve any CAPTCHA / 2FA.
  6. Click Save. The session is captured, encrypted, and stored.

How to use a session in a task

When creating a local or scheduled task, pick the saved session from the Browser session dropdown. The agent's BrowseWeb calls inside that task will run with your saved cookies / localStorage.

Expiry

If the session expires (cookie rotation, password change, etc.) the next task will fail with AUTH_SESSION_EXPIRED, the session will be marked Expired in the settings list, and a comment will be posted on the task notifying you. Click Re-login in the Browser Sessions list to capture a fresh state.

Security

  • Sessions are encrypted with a personal key derived per user. Other users cannot read them. Admins can revoke and delete sessions, but cannot decrypt them.
  • Sessions are not shared with org / public visibility — they are always bound to the task owner.
  • Audit logs record every save / use / decrypt with timestamp, actor, and result.

Limitations (v1)

  • Sessions are read-only snapshots — cookie mutations during a task run are NOT written back. Sites that rotate refresh tokens on every request may need re-login periodically.
  • IndexedDB and sessionStorage are not captured by Playwright.context.storageState, so sites that depend heavily on them may not work.
  • One profile, one site. Cross-domain SSO sessions need every involved origin visited during the initial login.