# mcpindex.ai

The agent-native index of Model Context Protocol (MCP) servers.

Built for IDEs and agents that find the right MCP server at inference time, not the developer browsing a sidebar.

## Scale

- Servers indexed: 7318
- Categories: 29
- Source: registry.modelcontextprotocol.io (canonical), enriched with quality scoring and semantic search.

## Endpoints an agent can call

- GET /api/v1/search?q=<query>           Keyword + semantic search across servers.
- GET /api/v1/recommend?task=<text>      Natural language task -> top 3 servers with reasoning.
- GET /api/v1/diff?since=<YYYY-MM-DD>    What changed in the registry since a date.
- GET /api/registry-count                Live server + category count.
- GET /llms-full.txt                     Full per-server index in one document.
- GET /.well-known/mcp-index.json        Machine-readable site capability descriptor.

## MCP server (drop-in)

```bash
npm install -g mcp-server-mcpindex
```

Add to Claude Desktop / Cursor / Cline / Zed and call `recommend_mcp_for_task("read pdfs and write to s3")`.

## Project pages

- /docs                     How it works, how to wire it into Claude/Cursor/Cline/Zed, response anatomy.
- /server/<slug>            Per-server detail (3,000+ pages, JSON-LD typed).
- /best/<category>          Curated picks per category.
- /leaderboard              Top 50 by MCP Quality Score.
- /changelog                Weekly diff of registry changes.
- /changelog.rss            RSS 2.0 feed of the above.
- /methodology              Open quality-score methodology.
- /about                    Project description + author.

Unofficial. Not affiliated with Anthropic.