browserbase_skill

Overview

This skill provides cloud browser infrastructure for AI agents and automation, letting you create and manage remote browser sessions, persist authentication state, and retrieve session artifacts. It exposes APIs to start Playwright/Puppeteer or Selenium sessions, debug live sessions, and collect logs, recordings, and downloads. Use it to offload browser execution and maintain reproducible authenticated contexts across runs.

How this skill works

The skill issues authenticated API calls to create sessions, list and update session state, and fetch artifacts like rrweb recordings, logs, and downloaded files. Sessions return connection URLs (WebSocket, Selenium) and signing keys for programmatic clients; contexts store cookies, cache, and storage that can be reused by subsequent sessions. Debug endpoints provide live DevTools and WebSocket URLs for real-time inspection while a client is connected.

When to use it

• Run headless/browser automation for scraping, testing, or agent-driven workflows without local browser infrastructure.
• Persist and restore login state across runs using contexts to avoid repeated authentication.
• Debug flaky flows in real time with live debugger URLs and session recordings after the run.
• Collect session logs, downloads, and rrweb replays for audit or troubleshooting.
• Manage session lifecycle programmatically to avoid overage charges by releasing sessions early.

Best practices

• Set BROWSERBASE_API_KEY and BROWSERBASE_PROJECT_ID in environment variables and use bash -c when piping to avoid variable loss.
• Use contexts to persist authenticated state and set persist: true to save updates after a session.
• Choose a region near target sites and enable proxies only on paid plans when needed for geo or rate-limit workarounds.
• Set sensible timeouts and keepAlive:true for longer sessions; explicitly request release to stop billing early.
• Streamline queries by writing complex filters to files and using --data-urlencode to avoid shell escaping issues.

Example use cases

• Automated login and data extraction from sites that require persistent cookies across runs.
• End-to-end UI test runs using Playwright/Puppeteer connected via the provided WebSocket URLs.
• Human-in-the-loop debugging of a bot session via the debuggerFullscreenUrl and live pages list.
• Forensic troubleshooting by downloading session logs, rrweb recordings, and files retrieved during a session.
• Running geographically distributed browsing via region selection and paid proxy support.