Model Context Protocol

Connect your AI coding tool

The MCP endpoint exposes one tool, skill-catalog, which returns the most relevant skills for a task prompt. Inline arguments such as --domain, --tags, and --skills-limit are supported in the prompt.

Endpoint

/api/mcp

HTTP transport. No authentication is required for the initial release.

MCP config

The example uses the current app host automatically, so it works for local development and deployed environments.

{
  "mcpServers": {
    "skill-catalog": {
      "url": "https://<host>/api/mcp",
      "transport": "http"
    }
  }
}

Tool input

Call skill-catalog with a prompt.

{
  "prompt": "Create a Payload hook --domain=Payload CMS --tags=hooks --skills-limit=3"
}

Supported arguments

--domain

Filters results toward a skill domain, such as Payload CMS or Cursor.

--tags

Filters by one or more comma-separated tags, such as hooks or deployment,evaluation.

--skills-limit

Sets the maximum number of skills to return. Defaults to 3.

Client setup examples

Any MCP client that supports Streamable HTTP can use this endpoint. These are examples for common tools.

Claude Code

Add the server block to your Claude Code MCP configuration, then restart Claude Code so it discovers skill-catalog.

Claude Code MCP docs

GitHub Copilot

Add the same server block to Copilot's MCP server settings for the workspace or user profile, depending on how your editor is configured.

VS Code MCP server docs

Cursor

Add the server block to .cursor/mcp.json for the project or ~/.cursor/mcp.json globally, then check MCP tools in Cursor settings.

Cursor MCP docs

PhpStorm / IntelliJ IDEA

Install or enable JetBrains AI Assistant, then open Settings/Preferences | Tools | AI Assistant | Model Context Protocol (MCP). Click Add, choose an HTTP or JSON MCP server configuration, paste the endpoint URL /api/mcp, choose the project or global level, and click Apply to start it.

JetBrains AI Assistant MCP docs

OpenCode

OpenCode MCP docs

Verify the endpoint

Use the MCP Inspector to confirm the public endpoint exposes skill-catalog. In the Inspector UI, choose Streamable HTTP and use /api/mcp as the server URL, or run the CLI command below.

npx @modelcontextprotocol/inspector --cli https://<host>/api/mcp --transport http --method tools/list