Decoupler

decoupler-mcp provides a natural language interface for scRNA-Seq analysis through MCP, enabling you to perform data IO, pathway activity and transcription factor inference, a broad toolset for clustering and differential expression, and plotting capabilities like violin and visualization with UMAP or t-SNE all via an accessible MCP client.

How to use

Connect to the MCP server from your MCP client and start issuing natural language prompts to drive scRNA-Seq analysis. You can run the server locally for development or deploy it remotely for production access. Use the http endpoint for remote access or the local stdio endpoint for direct integration with your tooling.

How to install

Prerequisites: ensure you have Python and pip installed on your system. You may also install from a Python environment manager if you prefer.

Install the MCP package from PyPI and verify the command is available on your system.

Test the installation by running the MCP entry point in your environment.

Extra configuration and run options

You can run the MCP server locally or remotely. Below are example configurations you can adapt in your MCP client to connect to either endpoint.

"mcpServers": {
  "decoupler_mcp": {
    "command": "/home/test/bin/decoupler-mcp",
    "args": [
      "run"
    ]
  }
}

Remote server usage details

To run the server remotely and expose an HTTP MCP endpoint, start the server with a transport option and port, then point your MCP client to the provided URL.

decoupler-mcp run --transport shttp --port 8000

"mcpServers": {
  "decoupler_mcp": {
    "url": "http://localhost:8000/mcp"
  }
}

Available tools

io_module

IO module for reading and writing scRNA-Seq data inside the MCP flow.

pathway_inference

Pathway activity and transcription factor inference capabilities.

tool_module

Tool module enabling clustering, differential expression, and related analyses.

plotting_module

Plotting utilities such as violin plots and dimensionality reduction visualizations (UMAP/TSNE).