decoupler-MCP

MCP server that exposes the Python decoupler footprint-analysis workflow as natural-language tools — infer pathway activities (PROGENy) and transcription-factor activities (CollecTRI) from single-cell or bulk expression.

   
Type MCP server
Supplier scmcphub
Availability GA — v0.4.0 (2025-06-27)
Pricing Free / OSS — BSD-3-Clause (per the scmcphub ecosystem; per-repo LICENSE file not separately confirmed)
Capabilities Read/Write — runs inference on a loaded AnnData object

How to install

Install the package first (required before any registration snippet below works):

pip install decoupler-mcp
  • Claude Code — stdio (recommended; Claude Code launches the process itself, so do not run decoupler-mcp run separately):
    claude mcp add --transport stdio decoupler -- decoupler-mcp run
    
  • Claude Desktop — add to claude_desktop_config.json:
    {
      "mcpServers": {
        "decoupler": { "command": "decoupler-mcp", "args": ["run"] }
      }
    }
    
  • HTTP transport (long-lived service — keep it running in its own terminal; not needed for the stdio setups above):
    decoupler-mcp run --transport shttp --port 8000
    

To verify the install before registering, run decoupler-mcp run once and Ctrl-C after it boots — for stdio the host launches it for you.

What it does

  • pathway_activity — infer pathway activities using PROGENy with the MLM (multivariate linear model) method
  • tf_activity — infer transcription-factor activities using CollecTRI with the ULM (univariate linear model) method

Primary use cases: Pathway-activity scoring per cell/sample, transcription-factor activity inference, functional interpretation of differential expression.

Notes

Built on decoupler. Part of the scmcphub single-cell ecosystem alongside scanpy-mcp, cellrank-mcp, and liana-mcp — typically run together on a shared AnnData object.

Sources


Installed this tool?

Share feedback — install path, OS, errors, workarounds. The form opens with this tool pre-selected and a link back to this page.