MCP

Interact with the Commonware Library via Model Context Protocol.

Why Provide an MCP Server?

LLMs are trained on code from months ago. Web search (the default fallback for finding missing information) returns GitHub links that must be iterated file-by-file to extract relevant information (if not rate-limited first). And, the results you do find probably don't match the version you're building against.

We built our own MCP server to make LLMs building with the Commonware Library more effective. mcp.commonware.xyz provides unlimited access to a version-pinned index of all source code and documentation, along with a ranked search tool that surfaces more relevant snippets than grep (with surrounding context).

Claude Code

claude mcp add --transport http commonware-library https://mcp.commonware.xyz

Cursor

Add to ~/.cursor/mcp.json (global) or .cursor/mcp.json (project):

{
  "mcpServers": {
    "commonware-library": {
      "url": "https://mcp.commonware.xyz"
    }
  }
}

Other Clients

Use the Streamable HTTP endpoint: https://mcp.commonware.xyz

If your AI assistant does not support MCP, you can use llms.txt to provide context manually.

Try It