what is browser-mcp fr?

This repository is a Model Context Protocol integration that gives Claude Desktop the ability to control a web browser on its own. It connects Claude to a real browser through Playwright, so the AI can open web pages, click buttons, fill in forms, take screenshots, and run JavaScript inside the page, all triggered by plain language instructions instead of code. Beyond browser control, the project also includes an API client tool. This lets Claude send HTTP requests like GET, POST, PUT, PATCH, and DELETE, set custom headers and request bodies, and read back the response data in a readable JSON format, with clear error messages when something goes wrong. The server also exposes browser console logs and screenshots as MCP resources, meaning Claude can retrieve them directly rather than needing a separate step. Because the browser session stays open between actions, Claude can chain together several steps, such as searching a site, clicking a result, and then filling out a form, while recovering from small errors along the way. To use it, a person needs Node.js 16 or later and Claude Desktop installed. Setup involves cloning the repository, installing dependencies with npm, and building the project, after which Playwright downloads the browser engines it needs for Chrome, Firefox, Edge, and WebKit, which stands in for Safari. An optional Redis connection can be added to cache API responses for 60 seconds and limit repeated requests to 60 per minute per host, though the tool works fine without it. The included demo video walks through real examples: searching Google, taking a screenshot, searching Wikipedia, picking options from a dropdown menu, and filling out and submitting a login form. The README does not mention a license, so its usage terms are not stated. The full README is longer than what was shown.