MCP – Project API

In a project that uses Vitek, you can expose that project's API (routes, sockets, manifest, OpenAPI, AsyncAPI) to AI assistants via Model Context Protocol (MCP). The AI helping with the front-end or another service can then see endpoints and types without you pasting documentation manually.

Command

From your project directory (where vite.config and the src/api folder live), run:

vitek mcp

This starts an MCP server over stdio. The server uses the current directory as the project root and reads the optional config from vitek.mcp.json.

Configuration (optional)

Create vitek.mcp.json at the project root to adjust paths and base URL:

{
  "apiDir": "src/api",
  "apiBasePath": "/api",
  "socketBasePath": "/api/ws",
  "baseUrl": "http://localhost:5173"
}

Field	Default	Description
apiDir	src/api	Directory for routes and sockets
apiBasePath	/api	HTTP API path prefix
socketBasePath	/api/ws	WebSocket path prefix
baseUrl	http://localhost:5173	Base URL used by the vitek_api_call tool

If the file is missing, these defaults are used.

Resources (read-only)

The server exposes the following resources:

URI	Content
vitek-api://manifest	Full manifest (routes, middlewares, sockets)
vitek-api://routes	HTTP routes list (method, pattern, params, file)
vitek-api://sockets	Sockets list (pattern, params, file)
vitek-api://openapi	OpenAPI 3.0 spec generated from routes
vitek-api://asyncapi	AsyncAPI 2.x spec generated from sockets

The AI can read these resources to discover your API surface.

Tool (optional)

• vitek_api_call – Calls a local API endpoint. Parameters: method, path (relative to apiBasePath), body (optional), headers (optional).
  Requirement: the API must be running (e.g. pnpm dev or pnpm start). The base URL is the one set in baseUrl in the config (or http://localhost:5173).

Cursor

1. In a Vitek project, create or edit .cursor/mcp.json (or Settings → MCP).
2. Add the server pointing to the vitek mcp command in the project directory:

{
  "mcpServers": {
    "vitek-api": {
      "command": "pnpm",
      "args": ["exec", "vitek", "mcp"]
    }
  }
}

If you use npx instead of pnpm:

{
  "mcpServers": {
    "vitek-api": {
      "command": "npx",
      "args": ["-y", "vitek-plugin", "mcp"]
    }
  }
}

3. Restart Cursor or reload MCP. The AI will then be able to read the vitek-api://* resources and, if the API is running, use the vitek_api_call tool.

Claude Desktop

In the Claude config file (e.g. ~/Library/Application Support/Claude/claude_desktop_config.json on macOS), add:

{
  "mcpServers": {
    "vitek-api": {
      "command": "pnpm",
      "args": ["exec", "vitek", "mcp"]
    }
  }
}

Start Claude Desktop from your Vitek project directory (or set cwd as per Claude’s documentation).

Summary

• Command: vitek mcp (from the project directory).
• Config: vitek.mcp.json (optional).
• Resources: manifest, routes, sockets, openapi, asyncapi.
• Tool: vitek_api_call (requires the API to be running).
• Usage: configure the MCP client (Cursor, Claude Desktop) to run vitek mcp in the project cwd.