RepoRadar Now Has an MCP Server for AI Coding Agents

If you're working with an AI coding agent and it needs to check a repository's star history, figure out what a project actually does, or see what's trending today, it's had exactly one option until now: fetch a page and re-parse the HTML. That works, but it's slow and it throws away all the structure RepoRadar already computed.

We just added a Model Context Protocol server at /api/mcp that exposes the same data as five queryable tools instead:

• search_repos — search tracked repositories by name, topic, or keyword
• get_repo — a repo's full profile: stars, AI-generated summary, use cases, getting-started command
• get_trending — what was trending on GitHub for any tracked day
• browse_category — repos by category, or a list of all categories
• get_report — a published daily, weekly, or monthly report

All five are read-only, and every one is backed by the exact same data functions the site's own pages use — nothing is served differently to agents than to a browser.

Adding it to your agent

If your client supports remote MCP servers over Streamable HTTP:

{
  "mcpServers": {
    "reporadar": {
      "url": "https://reporadar.spreix.com/api/mcp"
    }
  }
}

For Claude Code specifically:

claude mcp add --transport http reporadar https://reporadar.spreix.com/api/mcp

Full setup instructions, including a bridge for stdio-only clients, are on the MCP page.

Why build this instead of just better docs

A page is built for a human scrolling and clicking. An agent mid-task doesn't want to render a page, find the right section, and extract a number from prose — it wants a structured answer to a structured question. Exposing the same underlying data as typed tools instead of just HTML means an agent can ask "what does anthropics/claude-code actually do and how do I get started?" and get back exactly that, with nothing to parse.

This doesn't replace the site — most of what RepoRadar tracks (full star curves, contributor graphs, release history) is still there in the repo profiles for a human to explore. The MCP server is specifically for the moment an agent needs one fact fast, in the middle of doing something else.