@matrajs/mcp

The Matra documentation as a Model Context Protocol server, so any AI tool can read it. Zero dependencies, like every other Matra package.

npx -y @matrajs/mcp            # stdio · what a desktop client spawns
npx -y @matrajs/mcp --http     # http://localhost:3333/mcp

Connect it, step by step

Claude Code

claude mcp add matra -- npx -y @matrajs/mcp

Claude Desktop — in claude_desktop_config.json:

{
  "mcpServers": {
    "matra": { "command": "npx", "args": ["-y", "@matrajs/mcp"] }
  }
}

Cursor — in .cursor/mcp.json, the same object under "mcpServers".

Codex — in ~/.codex/config.toml:

[mcp_servers.matra]
command = "npx"
args = ["-y", "@matrajs/mcp"]

Anything that speaks HTTP — run npx -y @matrajs/mcp --http 3333 and point the client at http://localhost:3333/mcp.

Then ask the tool something about Matra. It will call search_docs, read the page it needs, and answer from the documentation rather than from memory.

What it serves

| Tool | Does | |---|---| | list_docs | Every page, with its slug and a one-line description. | | read_doc { slug } | One page, as Markdown. | | search_docs { query, limit? } | Ranked pages with a snippet each. |

Every page is also a resource at matra://docs/<slug>.

The pages are the repository's Markdown — README, the engine notes, benchmarks, security, the changelog — and every page of matrajs.com/docs, converted to Markdown at build time and shipped inside the package. Nothing is fetched at runtime.

Use it from code

import { createServer } from '@matrajs/mcp'

const server = createServer(docs)
server.handle({ jsonrpc: '2.0', id: 1, method: 'tools/list' })

createServer is the protocol without a transport: one message in, one reply out. Put it behind whatever transport you already have.