playwright-mcp
Playwright MCP is Microsoft's official MCP server for browser automation. Instead of sending screenshots to a vision model, it gives the LLM structured accessibility snapshots of the page, so the model can click, type, fill forms and navigate based on the page's actual structure. That makes actions faster and more deterministic than pixel-based approaches.
It suits agents that need to work with real websites: exploratory testing, self-healing end-to-end tests, form filling, and long-running autonomous workflows that keep browser state. It supports Chrome, Firefox, WebKit and Edge, can keep a persistent login profile, run isolated sessions, or connect to your existing browser through an extension, and optional capabilities add vision, PDF and DevTools tools. The Playwright team itself points out that coding agents may be better served by Playwright CLI with skills, which uses fewer tokens; MCP remains the better fit when an agent needs persistent state and rich page introspection.
Getting started: requires Node.js 18+. Add the standard config (`npx @playwright/mcp@latest`) to VS Code, Cursor, Claude Desktop or another MCP client, or run `claude mcp add playwright npx @playwright/mcp@latest` in Claude Code. A Docker image is available for headless Chromium.
Features
Why choose it
Trade-offs
Maintenance check
- Last commit
- 2026-10-22
- Commits, last 90 days
- 35
- Releases, last 90 days
- 5
- Latest release
- v0.0.83 (2026-09-28)
- Open issues
- 1
- Recent issues with at least one comment
- 18/28
Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.
Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.
Most discussed open issues
- #1802 Playwright MCP / isolated Edge0 comments · 2026-10-24
Issue titles are quoted from GitHub. Open the link to check whether a problem still applies to the version you use.
Who it fits
- You want an agent to drive a real browser through structured accessibility snapshots instead of screenshots, so no vision model is needed.
- You need Firefox, WebKit or Edge as well as Chrome: the README's --browser option lists chrome, firefox, webkit and msedge.
- You want to test or script flows with isolated in-memory profiles (--isolated), device emulation (--device) or saved Playwright traces (--save-trace).
- Your main goal is debugging or performance analysis of a page; Chrome DevTools MCP exposes performance traces and network inspection as dedicated tools.
- You need proxies, CAPTCHA handling or large-scale scraping of protected sites; that is outside what the README claims for this server.
- You want to treat it as a sandbox: the project says its origin allow and block lists are not a security boundary.
How it differs
vs chrome-devtools-mcp★ 54.3kBoth are Apache-2.0 MCP servers that let an agent drive a browser. Playwright MCP is built for automation and supports Chrome, Firefox, WebKit and Edge; Chrome DevTools MCP is Chrome-only but adds performance traces, network analysis and console inspection from DevTools. Pick Playwright for cross-browser flows and testing, Chrome DevTools for debugging a page.
Playwright MCP runs a browser on your own machine and needs no account. The Browserbase server runs browsers in Browserbase's cloud and needs a Browserbase API key plus a model key; its GitHub repository is also archived as of 2026-10-15.
Compatibility
Quick start
Use cases
Alternatives
Featured in these lists
Comments
No comments yet. Be the first!