LABSCO SUMMARY

Scanpy's structure, kept intact.

Wrapping an analysis library usually means inventing a new vocabulary on top of it; here the modules stay where they were, so someone who knows Scanpy can predict what is available and someone who does not can be walked through a standard workflow without learning the API first. The remote transport is the part worth planning around: single-cell data is large, and running the server where the data lives rather than shipping the data to your laptop is what makes this practical.

What it is

A natural-language front end to Scanpy, the standard Python toolkit for single-cell RNA sequencing analysis. It mirrors Scanpy's own module structure, so the steps an analysis takes are the steps you already know — they are just reachable by asking rather than by writing the call.

What you get

The IO module: reading and writing single-cell datasets
The preprocessing module — filtering, quality control, normalization, scaling, highly-variable gene selection, PCA and neighbour graph construction
The tools module, covering clustering and differential expression
The plotting module: violin plots, heatmaps and dot plots
The same surface whether you drive it from an AI client, an editor plugin or an agent framework — the README names Cherry Studio, Cline and Agno as working hosts

Requirements

No account and no key. Install scanpy-mcp from PyPI, then either point your client at the installed executable with its run subcommand for a local stdio setup, or start it on a server with a streamable-HTTP transport and a port and connect the client to that URL. Documentation for the project lives at the maintainers' docs site.

Setup effort

One command — pip install scanpy-mcp

See installation options

USE CASES

01

Taking a dataset from raw counts to clusters

Walk the standard path — filter, QC, normalize, select variable genes, PCA, neighbours, cluster — by describing each step rather than writing it.

02

Finding what distinguishes a cluster

Run differential expression between groups and read the markers back, in the same conversation where you decided which groups to compare.

03

Looking at the data before deciding anything

Generate violin plots, heatmaps or dot plots for the genes in question, without leaving the analysis to write plotting code.

04

Giving an agent access to real single-cell tooling

Run it on the machine that holds the data over HTTP, so the analysis happens next to the dataset rather than on the client.

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": {
    "scanpy-mcp": {
      "command": "scanpy-mcp",
      "args": [
        "run"
      ]
    }
  }
}

If your client cannot find scanpy-mcp, put the full path printed by which scanpy-mcp in "command"

pip

TERMINAL

pip install scanpy-mcp

Scanpy-MCP