mult-fetch

mult-fetch-mcp-server implements an MCP-compliant client and server to facilitate communication between AI assistants and external web tools. It provides a modular fetching pipeline with multiple fetchers (browser-based, Node-based, and common utilities) and supports intelligent content extraction, page metadata, and content chunking to work within model context limits. The server is designed to run alongside an MCP client and communicates via standard IO (stdio) with optional SSE channels for message transport. To use it, configure your MCP client to launch this server as a child process and point it to the same MCP protocol endpoints. You can leverage its built-in tools for web scraping, content processing (HTML-to-text, metadata extraction), and chunked content retrieval to retrieve and process large pages in smaller, context-friendly segments.

Prerequisites:

• Node.js (recommended latest LTS) and npm/pnpm
• Internet access to install packages

Local installation (no global install required):

pnpm install

Run directly with npx (no prior installation required):

npx @lmcc-dev/mult-fetch-mcp-server

If you prefer a global installation:

pnpm add -g @lmcc-dev/mult-fetch-mcp-server

Then run the server with:

mult-fetch-mcp-server

Optionally, you can configure the server language via environment variable when launching from the client:

MCP_LANG=en npx @lmcc-dev/mult-fetch-mcp-server

Notes and tips:

• The server follows the MCP Stdio transport model; ensure your MCP client is configured to start the server as a child process and to communicate via stdin/stdout.
• You can adjust language and other behavior using environment variables (e.g., MCP_LANG).
• The server emphasizes intelligent content extraction and chunked content delivery to help models stay within context limits.
• If integrating with tools like Claude Desktop, use the provided configuration examples to specify how to start the server (npx, full path, or npm package path).
• Typical issues may include version mismatches between MCP SDK and the server build; ensure you’re using a compatible MCP SDK version (noted in the README as MCP SDK 1.7.0).