@directive-run/knowledge-mcp

A Model Context Protocol server that exposes the Directive knowledge package as MCP tools — every knowledge doc, code example, and Claude Code skill bundle, queryable at retrieval time instead of bundled as a static snapshot.

Two transports ship in the same binary:

• stdio — the canonical MCP local-client pattern. Plugs into Claude Desktop, Cursor MCP, the MCP Inspector, and any other client that spawns an MCP server as a subprocess.
• SSE (HTTP) — for hosting at mcp.directive.run or a private MCP gateway. Production AI agents query the hosted endpoint at runtime; no embedding pipeline, no bundle bloat.

Install

npm install -g @directive-run/knowledge-mcp
# or run directly without installing:
npx @directive-run/knowledge-mcp --help

stdio transport (local clients)

Add to your Claude Desktop config (claude_desktop_config.json):

{
  "mcpServers": {
    "directive": {
      "command": "npx",
      "args": ["-y", "@directive-run/knowledge-mcp"]
    }
  }
}

Or test with the MCP Inspector:

npx @modelcontextprotocol/inspector npx -y @directive-run/knowledge-mcp

SSE transport (hosted)

directive-knowledge-mcp --sse --port 3000 --host 0.0.0.0

Endpoints:

• GET /sse — establish the SSE stream.
• POST /messages?sessionId=… — client→server JSON-RPC messages.
• GET /healthz — liveness probe.

Tools

| Tool | Purpose | |---|---| | list_knowledge | Every knowledge file name (core + AI + skeleton). | | get_knowledge | Read one knowledge file by name. | | list_examples | Every code example name. | | get_example | Read one example by name (returned as a TypeScript code block). | | search_knowledge | Case-insensitive substring search across every knowledge file. | | list_skills | Every Claude Code skill bundled in @directive-run/claude-plugin. | | get_skill | One skill's SKILL.md + supporting knowledge files as a single document. |

Programmatic embedding

For tool authors who want to mount the server inside their own host process:

import { createDirectiveKnowledgeServer, startSseServer } from "@directive-run/knowledge-mcp";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";

// stdio
const server = createDirectiveKnowledgeServer();
await server.connect(new StdioServerTransport());

// SSE — returns the underlying http.Server
const httpServer = await startSseServer({ port: 3000, host: "0.0.0.0" });

See also

• @directive-run/knowledge — the knowledge package this server fronts.
• @directive-run/claude-plugin — the Claude Code skill bundles also exposed.
• @directive-run/cli — generate static .cursorrules / CLAUDE.md / .windsurfrules files for assistants that don't speak MCP.
• docs/ide-integration — decision tree across every install path.