QuDI

You can control qudi quantum photonics experiments through a dedicated MCP server that enables natural language control via Claude Desktop. This server provides instrument access, safety runlevels, measurement execution, and optional plot extraction, all orchestrated through a clear, scriptable interface that you can test safely before using real hardware.

How to use

You interact with the MCP server through an MCP client (Claude Desktop) by starting the server locally and then issuing natural language commands. Typical workflows include listing available instruments, loading an instrument, setting validated parameters, starting measurements, and reviewing results. Safety controls enforce runlevel transitions and emergency stops so you can test in a dry-run or simulated environment before connecting to real hardware.

How to install

Prerequisites: you need Python installed on your system. You also may want Claude Desktop for control via natural language commands.

Follow these concrete steps to set up a standalone MCP server for testing within a safe, simulated environment.

# Clone the MCP integration repository
git clone https://github.com/dirkenglund/qudi-mcp-integration.git
cd qudi-mcp-integration

# Install minimal dependencies for simulation only
pip install -r requirements-standalone.txt

# Optional: install with plot extraction capabilities
pip install -r requirements-standalone.txt
pip install -r requirements-plot-extraction.txt

# Optional: full integration for hardware control (requires additional setup)
pip install -r requirements-full.txt

# Run the MCP server (example command, use your actual path)
python qudi_mcp_server.py

Configuration and test with Claude Desktop

Configure Claude Desktop to connect to your local MCP server by adding a client configuration that launches the MCP server using Python, and sets the appropriate environment variables for logging and Python paths.

{
  "mcpServers": {
    "qudi-mcp": {
      "command": "/opt/homebrew/bin/python3",
      "args": ["/path/to/qudi-mcp-integration/simple_mcp_server.py"],
      "env": {
        "PYTHONPATH": "/path/to/qudi-mcp-integration",
        "QUDI_MCP_LOG_LEVEL": "INFO"
      }
    }
  }
}

Test with Claude Desktop

Restart Claude Desktop and say or type a simple query to verify the connection and capabilities, for example: “Get qudi station information”. You should see the current runlevel, loaded instruments, active measurements, and safety status.

Usage examples you can try

• Start an instrument operation by loading an instrument and setting a parameter safely.

• Run a PL scan or a time trace to observe data acquisition, with safety checks ensuring parameters stay within limits.

Available tools

instrument_tools

Tools to list, load, and configure qudi instruments with safety validation and status monitoring.

measurement_tools

Tools to start, monitor, and retrieve results from measurements such as PL scans, gate sweeps, and time traces.

safety_tools

Safety and emergency tools to check interlocks, set runlevels, and trigger emergency stops.

plot_extraction_tools

Optional tools for extracting data from plots and performing RKHS-based smoothing and spectrum analysis.