An MCP server that lets LLMs automate browser tasks through Puppeteer-based structured page data, with optional screenshot/vision mode for cases where text and accessibility data are not enough. It supports local, remote (SSE/HTTP), and in-memory use, and can connect to local or remote browsers.
From the registry: MCP server for browser use access
Please install the `mcp-server-browser` MCP server into my current AI client (that's you).
Required prerequisites (do these first if not already done):
- **Install Node.js 18 or newer** — Node.js version 18 or newer is required to run the MCP server via npx. (https://nodejs.org/.)
Canonical MCP server config (stdio transport):
- command: `npx`
- args: ["@agent-infra/mcp-server-browser@latest"]
Note: Primary install path is local stdio via npx. The server can also be started in remote mode by adding --port <port> to the args, which exposes Streamable HTTP at http://<host>:<port>/mcp and SSE at http://<host>:<port>/sse. Optional CLI args supported in stdio or remote mode include: --browser <browser>, --cdp-endpoint <endpoint>, --ws-endpoint <endpoint>, --executable-path <path>, --headless, --host <host>, --port <port>, --proxy-bypass <bypass>, --proxy-server <proxy>, --user-agent <ua string>, --user-data-dir <path>, --viewport-size <width,height>, --output-dir <path>, and --vision. For remote mode, runtime configuration may also be supplied via HTTP headers: x-viewport-size, x-vision-factors, and x-user-agent. DISPLAY may optionally be set in the shell when launching the server for VNC or virtual display usage, but no config-file environment variables are required by default.
Add this MCP server to my current client's config in the correct format for you. If you need secrets or credentials I haven't provided, ASK me — do not invent values or leave raw placeholders. After adding it, tell me how to verify the server is connected.MCP server for Firecrawl web scraping, structured data extraction and web search integration.