Web Browser
The universal interaction surface. Agent browser access splits between vision-based control (Browser Use), structured automation (Playwright), and AI-native SDKs (Stagehand).
6
Skills
1
Categories
Native agent support
Browsers have no native agent API. All agent access goes through automation libraries, MCP wrappers, or vision-based screenshot approaches.
Skills on this platform
Browser Use
activeBest default for agents that need to see and interact with real web pages end-to-end.
Community · browser-use/browser-use
Chrome DevTools MCP
activeBest MCP-native option for debugging, performance analysis, and Chrome-specific development workflows. Complements Playwright MCP rather than replacing it.
Official · ChromeDevTools/chrome-devtools-mcp
Playwright MCP
activeBest MCP-native browser option for teams in Microsoft's ecosystem. The accessibility-snapshot approach is more reliable than vision-based alternatives for structured data extraction.
Official · microsoft/playwright-mcp
Vercel Agent Browser
activeBest pick when token efficiency is the primary constraint — long autonomous sessions, cost-sensitive deployments, or agents hitting context limits.
Community · vercel-labs/agent-browser
Stagehand
activeBest pick when the team wants TypeScript-native browser automation with the simplest possible API surface.
Official · browserbase/stagehand
Skyvern
activeBest pick for enterprise workflow automation on websites without APIs — form filling, data entry, procurement. Overkill for developer/coding agent browser tasks.
Community · Skyvern-AI/skyvern
See how these skills rank.
Category-level rankings with evidence and head-to-head comparisons.
Related categories