Browser Automation

microsoft

Cross-browser automation — navigate, click, type, screenshot, and inspect live pages via the official Playwright MCP server.

Add to your mcp.json

{
  "mcpServers": {
    "playwright": {
      "command": "npx",
      "args": [
        "-y",
        "@playwright/mcp"
      ]
    }
  }
}

mcp_scope:
  attached:
    - playwright

none

no credentials required; first run downloads browser binaries on demand.

No credentials required.

Variable	Required	Auth mode	Description
PLAYWRIGHT_BROWSERS_PATH	no	none	Optional override for the browser-binary install location. Defaults to the Playwright cache dir; set to pin browsers in a shared/CI cache.

playwright.browser_navigate
Navigate the active page to a URL.
playwright.browser_click
Click an element by accessible selector.
playwright.browser_type
Type text into a focused input.
playwright.browser_snapshot
Capture an accessibility snapshot of the page.
playwright.browser_take_screenshot
Capture a screenshot of the page or element.

FAI cookbook recipes that attach playwright.

No destructive tools — audited read-only

Destructive-action audit (X2.18): no destructive tools — browser automation does not delete server-side resources.

First use downloads browser binaries (Chromium/Firefox/WebKit) — the initial attach may be slow and requires network + disk.
Headless by default; some sites that block automated browsers may behave differently than a real user session.

Maintainers who landed commits onmcp-specs/playwright.json.

When this spec was last reviewed by a maintainer and last machine-validated against the live server.