# Playwright MCP

> Lets AI agents automate browsers via accessibility snapshots

- Page on StartupList: https://startuplist.tech/launches/playwright-mcp
- Website: https://github.com/microsoft/playwright-mcp
- Category: MCP Servers
- Price: Open source
- Company: Microsoft (https://startuplist.tech/companies/microsoft-5)
- Used for: Automate workflows

## At a glance

- The problem it solves: Letting AI agents operate web pages without screenshots or vision models, using structured page data instead

## About

Playwright MCP is an open source Model Context Protocol server that gives language models browser automation through Playwright. It runs locally through npx and works with MCP clients such as VS Code, Cursor, Claude Desktop, Claude Code, Codex and many others.

Instead of relying on screenshots, the server reads the accessibility tree of each page, a structured snapshot of its elements. The project says this keeps the approach fast and lightweight and removes the need for vision models, so a model can act on pages using structured text alone.

The README also covers when to use it. It notes that coding agents may prefer Playwright's CLI with skills because CLI calls use fewer tokens. It says MCP suits longer agentic loops that need persistent browser state and iterative reasoning over page structure, such as exploratory automation or self-healing tests.

Setup requires Node.js 18 or newer. Most clients use the same JSON entry, which starts the server with npx @playwright/mcp@latest. Configuration options cover browsers such as Chrome, Firefox, WebKit and Edge, allowed and blocked origins, file access limits, and the language used for code generation.

The package is licensed under Apache-2.0 and its source is on GitHub. The page does not state a price for the server, so the listing shows it as open source.

## Links

- [Playwright MCP on StartupList](https://startuplist.tech/launches/playwright-mcp)
- [Website](https://github.com/microsoft/playwright-mcp)