Connecting AI Clients

The TanStack CLI MCP server runs locally via stdio (default) or HTTP/SSE.

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):

{
  "mcpServers": {
    "tanstack": {
      "command": "npx",
      "args": ["@tanstack/cli", "mcp"]
    }
  }
}

Windows: %APPDATA%\Claude\claude_desktop_config.json

Restart Claude Desktop after editing.

claude mcp add tanstack -- npx @tanstack/cli mcp

Add to your Cursor MCP config:

{
  "mcpServers": {
    "tanstack": {
      "command": "npx",
      "args": ["@tanstack/cli", "mcp"]
    }
  }
}

Add to .vscode/mcp.json or VS Code settings:

{
  "servers": {
    "tanstack": {
      "command": "npx",
      "args": ["@tanstack/cli", "mcp"]
    }
  }
}

For any MCP client supporting stdio transport:

• Command: npx
• Args: ["@tanstack/cli", "mcp"]

Or with global install:

• Command: tanstack
• Args: ["mcp"]

For clients preferring HTTP:

tanstack mcp --sse

Connect to http://localhost:8080/sse.

Test the server interactively:

npx @modelcontextprotocol/inspector -- npx @tanstack/cli mcp

After connecting, ask your AI assistant:

"List all available TanStack libraries"

You should see output from tanstack_list_libraries showing Query, Router, Table, and other libraries.