What does Browser JavaScript Evaluator do?
It allows you to execute JavaScript code in a browser from Python via MCP. A web server sends code to a connected browser using SSE, and the browser returns results.
@matdmiller
基本信息
This is a reference design for a MCP server that hosts a web page that connects back to the server via sse and allows Claude to execute javascript on the page.
使用下面的配置,将此服务器添加到你的 MCP 客户端。
{
"mcpServers": {
"mcp_py_sse_browser_control": {
"command": "python",
"args": [
"web_server.py"
]
}
}
}工具列表暂未提供
当前收录暂时没有可展示的工具列表,不代表服务没有工具。请查阅项目文档了解支持的能力。
工具是从 README 中自动提取的。维护者可以在 ## Tools 标题下列出工具,即可填充这部分内容。
Browser JavaScript Evaluator is an MCP (Model Control Protocol) server that lets you execute arbitrary JavaScript code in a browser from Python. It uses a web server with SSE (Server-Sent Events) and a REST API to relay code to a connected browser and return results.
Add the server to your Claude Desktop configuration using uv pointing to the repository directory and the browser_server.py script. After launching Claude, open a browser to http://127.0.0.1:8000 to connect. The server provides two tools: execute_javascript (run any JavaScript) and add_numbers (add two numbers via browser execution).
It allows you to execute JavaScript code in a browser from Python via MCP. A web server sends code to a connected browser using SSE, and the browser returns results.
Add the server to your Claude config with the `uv` command pointing to the repository, then launch Claude. Open a browser at http://127.0.0.1:8000 to connect. Alternatively, run `python web_server.py` to start only the web server.
Two tools: `execute_javascript` (runs any JavaScript) and `add_numbers` (adds two numbers using JavaScript).
Yes. The server can execute arbitrary JavaScript in your browser, which is potentially dangerous. Use at your own risk and be aware of the implications.
The server runs with Python and uses the `uv` package manager. A browser must be open and connected to the web server (default at port 8000) for code execution to work.
Allow LLMs to control a browser with Browserbase and Stagehand
Browse the web, directly from Cursor etc.
Give your AI agent the live web. One MCP server for Search, Fetch, Browser, and Agent: browse real sites, extract structured data, fill forms, and run multi-step workflows. Search and Fetch are free.
MCP server for fetch web page content using Playwright headless browser.
Monitor browser logs directly from Cursor and other MCP compatible IDEs.
评论