rag

This MCP server is a CodeRAG-enabled runner. It provides a fast, dependency-free way to deploy a code-aware retrieval system that combines TF-IDF keyword signals with vector-based ranking for code search. When started via npx @sylphx/coderag-mcp, it boots CodeRAG’s MCP backend in a lightweight, local environment and exposes tools for indexing and querying code repositories. The system supports incremental indexing, hybrid search (TF-IDF + vector), code-aware tokenization, and optional SQL-backed low-memory mode. Use it to index your codebase and perform fast, accurate code search across languages and repositories. To interact with the MCP server, you’ll rely on the same MCP-style endpoints and queries you use with other MCP servers in your workflow, enabling easy integration with your existing tooling.

Prerequisites:

• Node.js (recommended latest LTS) and npm/yarn installed on your machine
• Internet access to fetch the MCP package

Installation steps:

1. Install Node.js if you don’t have it already. Visit https://nodejs.org/ and install the LTS release.

2. Ensure npm or yarn is available in your environment.

3. Run the MCP installer via npx (no global install required):

npx -y @sylphx/coderag-mcp

4. Follow the prompts (if any) to initialize the CodeRAG MCP server configuration. You can also explicitly specify the MCP config using your preferred orchestration, e.g. using the provided mcp_config:

// Example config segment (loaded by your MCP manager)
{
  "mcpServers": {
    "coderag": {
      "command": "npx",
      "args": ["-y", "@sylphx/coderag-mcp"]
    }
  }
}

5. Run your MCP configuration through your usual MCP runner (the exact command depends on your environment).

Notes and tips:

• The CodeRAG MCP does not require external dependencies like ChromaDB or Ollama, making it lighter and faster to start.
• You can index code repositories incrementally, enabling smart diffs rather than full rebuilds.
• If you need a low-memory setup, explore the SQL-based option mentioned by CodeRAG documentation.
• Ensure network access is available if you plan to fetch remote dependencies or models during indexing (even though this MCP emphasizes zero-dependency operation, some setups may rely on optional components).
• The legacy rag-server-mcp is deprecated; prefer coderag-mcp for new projects.
• If you upgrade, be mindful of configuration migrations: replace old server keys (e.g., rag-server) with coderag in your mcpServers configuration.