Decoupler MCP
scmcphubLABSCO SUMMARY
The whole infercnvpy path, without writing the notebook.
What this replaces is the boilerplate around an analysis you already know how to do: loading positions, computing neighbours, running inference, plotting. That makes it useful for exploring a dataset conversationally and for agent code that needs CNV inference as one step among several. It is a wrapper, so the underlying method, its assumptions and its limits are infercnvpy's, and worth reading there before you trust a heatmap.
What it is
An MCP server wrapping infercnvpy, the Python package for inferring copy number variation from scRNA-seq data. It covers the whole path in one place: reading and writing the data and loading gene positions, computing neighbours and preparing the object, running the CNV inference and scoring, and plotting the result.
What you get
- Data read in and written back out, with gene position information loaded — the IO module
- Neighbour computation and the preparation steps the inference expects — the preprocessing module
- CNV inference over the prepared object, plus a CNV score — the tool module
- Chromosome heatmaps and UMAP or t-SNE embeddings of the result — the plotting module
Requirements
No account and no key — it runs against data on your own machine. Install with pip install infercnv-mcp; infercnv-mcp run starts it over stdio, and a client entry points at the absolute path of the installed binary with the run argument. To reach it from another machine, infercnv-mcp run --transport shttp --port 8000 serves it at http://localhost:8000/mcp. The pyproject name is infercnv-mcp. Full documentation for the scmcphub family is at docs.scmcphub.org.
Setup effort
One command — pip install infercnv-mcp
USE CASES
01
Getting a first CNV read on a new dataset
Load the matrix and gene positions, run the preparation and inference, and look at the chromosome heatmap without writing a script.
02
Separating likely malignant cells from normal ones
Compute the CNV score across cells and view it on a UMAP or t-SNE embedding to see whether it tracks the clustering.
03
Adding CNV inference to an agent workflow
Run it remotely over streamable HTTP so an agent framework can call the analysis as one step in a longer pipeline.
BEFORE YOU INSTALL
What to know first
Requires
pip on PATH
INSTALLATION
Connect your client
Add the block to your client, or run the command directly.
Server setup
Configuration file
JSON
{
"mcpServers": {
"infercnv-mcp": {
"command": "infercnv-mcp",
"args": [
"run"
]
}
}
}If your client cannot find infercnv-mcp, put the full path printed by which infercnv-mcp in "command"
pip
TERMINAL
pip install infercnv-mcp
