Playwright MCP Server

Enables AI agents to automate browsers (Chrome, Firefox, Safari) through the Model Context Protocol using Playwright. Supports navigation, clicking, form filling, screenshots, and content extraction across multiple browser engines. More robust than Puppeteer for cross-browser testing scenarios.

Security Assessment

Level	caution
Permissions	browse:web, capture:screenshots, interact:pages
Data Exposure	content
Requires Supervisor	No
Risk Notes	Browser automation. Can navigate, click, fill forms, take screenshots.

Integration

Type	`mcp-server`
Package	`@playwright/mcp`
Auth	none

Capabilities

Actions	navigate_to_url, take_screenshot, click_element, fill_form, extract_content, select_option
Input Types	text, url
Output Types	text, image, json
When to Use	When the agent needs cross-browser automation, testing, or scraping with support for multiple browser engines.
When NOT to Use	When simple HTTP fetching is sufficient; use fetch-mcp for static content.

Category: content · Free tier: Yes · Source

Collected: 2026-03-15 from https://github.com/microsoft/playwright-mcp