MCP Slicer

You can connect 3D Slicer to model clients through the Model Context Protocol (MCP) to directly control image processing, scene creation, and visualization from natural-language or scripted interactions. This MCP server lets you send commands to Slicer, inspect its scene, run Python in the Slicer environment, and capture visual feedback for a complete AI-assisted workflow.

How to use

You use this MCP server by running it locally and connecting your MCP client (for example Claude Desktop or Cline) to the provided interface. Start the local MCP server, then issue actions from your client such as listing nodes in the Slicer scene, executing Python code inside Slicer, or capturing screenshots for visual feedback. The server enables a REACT-like loop where reasoning, acting, and observing guide the workflow.

How to install

Prerequisites include 3D Slicer 5.8 or newer, Python 3.13 or newer, and the uv package manager. Install uv first to ensure smooth startup.

# Install uv (example for macOS)
brew install uv

# On Windows, install uv
powershell -c "irm https://astral.sh/uv/install.ps1 | iex"
set Path=C:\Users\nntra\.local\bin;%Path%

MCP server configuration

Configure your MCP client to launch the slicer MCP server using the stdio method. The server runs as a local process and uses uvx to start the MCP pipeline.

{
  "mcpServers": {
    "slicer": {
      "command": "uvx",
      "args": ["mcp-slicer"]
    }
  }
}

Available tools

list_nodes

List and filter Slicer MRML nodes and view their properties to understand the current scene state.

execute_python_code

Run Python code inside the Slicer environment to manipulate data, create or modify nodes, and perform processing.

capture_screenshot

Capture real-time screenshots of Slicer views to provide visual feedback during an interaction loop.