Playwright

Why use it

Key features

• browser_wait_for accepts time, text or textGone; time is in seconds, at most 30
• Use --browser firefox or --browser webkit to select a browser
• Read a fresh snapshot before interacting with the page

Example replay

Illustrative saved transcript; historical tool calls and timings are not current API documentation or a live execution.

What it looks like in practice

Install

Claude Desktop / Cursor: local stdio

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

Requires Node.js 18+. Add --headless for headless operation or --browser firefox to select Firefox. Browser availability and OS dependencies vary by platform.

Client/version scope: Claude Desktop / Cursor. Other clients: pending verification.

Use Cases

Suggested prompts, not measured outcomes. Installation and listed tools were reviewed; these workflows were not executed.

Smoke-test a production deploy in 60 seconds

Open https://app.example.com. Verify the page returns 200, the H1 says 'Welcome', and the login button is visible.

Log in as [email protected] / [password]. Then create a new project named 'smoke-test-<timestamp>'. Then delete it. Then log out.

If a step failed, capture the page snapshot and tell me what's different from a working baseline.

Audit your site at 5 viewport widths and report breakage

Open https://example.com at viewports 320×568, 375×812, 768×1024, 1024×768, 1440×900. Take a full-page screenshot of each.

For each viewport, list any elements where scrollWidth > viewportWidth. These are causing horizontal scroll and feel broken.

For each overflow element, get its computed CSS and suggest the smallest fix (probably max-width, overflow-wrap, or breakpoint adjustment).

Crawl your site and find every broken internal link

Starting from https://docs.example.com, crawl all pages on the same domain (depth 3 max). Collect every href found.

For each unique URL, send a HEAD request and capture the status code. Flag anything >= 400.

For each broken link, list which page(s) link to it, so I know where to fix.

Scrape data that requires authentication

Open https://target.example.com/login. Fill the email field with [email] and password with [password]. Submit.

Click into the Reports section. Wait for the data table to render.

Read every row of the report table. Return as JSON with each column as a key.

Combinations

Integration compatibility: pending verification. Treat these as suggested workflows.

Crawl docs.stripe.com starting from /api, save each page as a Markdown file under ./knowledge/stripe/{path}.md.

Open my landing page. Verify the H1, then capture a Lighthouse trace and tell me if LCP is under 2 seconds.

Sentry issue WEB-3a91 says the user clicked X then Y then crashed. Reproduce that exact sequence in Playwright and confirm.

Tools

Selected parameters checked against upstream definitions. Consult the connected server tools/list for your installed version.

• browser_navigate

  url: string

  Open a page
• browser_snapshot

  target?: string, filename?: string, depth?: number, boxes?: boolean

  Inspect the current page
• browser_click

  target: string, element?: string, doubleClick?: boolean, button?: string, modifiers?: string[]

  Click a snapshot target or unique selector
• browser_type

  target: string, text: string, element?: string, submit?: boolean, slowly?: boolean

  Fill a field
• browser_wait_for

  time?: number (seconds, at most 30), text?: string, textGone?: string

  Wait for text or elapsed time

Cost & Limits

api_quota: pending verification

per_call_tokens: pending verification

monetary: pending verification: see current upstream pricing and hosting requirements.

budget_tip: Limit requested data and inspect actual output before expanding queries.

Security

Pending verification: client and browser credential handling.

Browser requests reach visited sites; a complete telemetry audit is pending review.

Allowed/blocked origins are not a security boundary according to the README.

CVE status and a full security audit: pending verification.

Troubleshooting

Wait or interaction parameter mismatch

Use browser_wait_for with text, textGone or time, not selector. Refresh the snapshot and use the advertised target parameter for click/type.

Browser selection fails

Use --browser firefox (not --browser-type=firefox). Check the upstream platform instructions for browser dependencies.

Alternatives

Pending verification: alternative projects and tradeoffs have not been rechecked.

Official sources

Documentation checked 2026-10-08. No authenticated service was executed. Unchecked claims are marked pending verification.

• https://github.com/microsoft/playwright-mcp — checked 2026-10-08

More

Resources

📖 Read the official README on GitHub

🐙 Browse open issues

🔍 Browse all 400+ MCP servers and Skills