Agent Connection¶

TwinCat-MCP speaks the Model Context Protocol over Streamable HTTP. Point any MCP-capable agent at the Service endpoint:

http://localhost:3001/mcp

Make sure the Service is installed and running first.

Claude Code¶

Install Claude Code if you haven't already:

npm install -g @anthropic-ai/claude-code

Add the MCP server:

claude mcp add --transport http twincat-mcp-service http://localhost:3001/mcp

Then start Claude Code in your TwinCAT project directory:

cd C:\MyProjects\PlcProject
claude

GitHub Copilot CLI¶

Install the Copilot CLI via winget:

winget install GitHub.Copilot

Note

Requires a GitHub Copilot subscription (Individual, Business, or Enterprise). Open a new terminal after installation for copilot to be available.

Create or edit %USERPROFILE%\.copilot\mcp-config.json:

{
  "mcpServers": {
    "twincat-mcp-service": {
      "type": "http",
      "url": "http://localhost:3001/mcp"
    }
  }
}

Verify with /mcp show inside the Copilot CLI session. Project-level configuration is also possible at .copilot/mcp-config.json in the project root (same format).

Cursor¶

Open Cursor Settings (Ctrl+Shift+J) → MCP section and add a new server:

Name: twincat-mcp-service
Type: http
URL: http://localhost:3001/mcp

Alternatively, add it to .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "twincat-mcp-service": {
      "type": "http",
      "url": "http://localhost:3001/mcp"
    }
  }
}

Next¶

Once connected, your agent lists every available tool with its description. See the Tools Overview for what's on offer, and Licensing to activate them.