language

The language-server-mcp provides language support features for code editing via the Model Context Protocol (MCP). It exposes tools such as get_hover, get_completions, and get_diagnostics to offer hover information, code completion suggestions, and diagnostics for TypeScript (with potential future support for Python). The server is implemented in TypeScript and when built by npm run build, it outputs a runnable JavaScript entry point (build/index.js). To use it, add the server to your MCP client configuration so that standard input/output channels are used for communication, enabling real-time language features within your editor or IDE integration.

Prerequisites:

• Node.js and npm installed on your system
• A working MCP client or environment to connect to MCP servers

1. Install dependencies npm install

2. Build the server (transpile TypeScript to JavaScript) npm run build

3. (Optional) Run in watch mode for auto-rebuild during development npm run watch

4. Run or configure your MCP client to connect to the server via the generated entry point

   • The built entry point is typically at build/index.js

If you plan to debug, ensure the MCP Inspector tooling is available and accessible as described in the README: npm run inspector

Notes and tips:

• The server communicates over standard I/O as part of the MCP protocol. When debugging, consider using the MCP Inspector tooling referenced in the README for easier inspection of requests and responses.
• The server currently emphasizes TypeScript language features (hover, completions, diagnostics) with potential expansion to other languages in the future.
• To integrate with Claude Desktop, you can point Claude’s config to the built entry at build/index.js. Ensure the path matches your environment.
• If you customize environment variables or paths, keep the MCP logging verbose during development to ease troubleshooting.