UCSC Cell Browser MCP

Find single-cell RNA-seq datasets in the UCSC Cell Browser by organism, body part, disease or project, and retrieve each dataset’s metadata without leaving Claude.


Type	MCP server
Supplier	MCPmed
Availability	Beta — `pyproject.toml` declares v1.0.0; last upstream commit 2025-07-28
Pricing	Free / OSS — BSD-3-Clause (repository `LICENSE`); note the `pyproject.toml` `license` field says MIT, so the two disagree
Capabilities	Read-only — queries the public `cells.ucsc.edu` dataset index; no account, no API key

How to install

The README’s pip install ucsc-cell-browser-mcp does not work — that package is not published on PyPI (404 as of 2026-08-15). Install from source instead.

Clone and install (the README’s cd ucsccbMCP is a typo; the clone directory is UCSCCBmcp):
```
git clone https://github.com/MCPmed/UCSCCBmcp
cd UCSCCBmcp
pip install -e .
```
This installs the ucsc-mcp console command (entry point ucsccbmcp.main:main_cli). Python 3.8+ required.
Find the absolute path to the installed command — you need it for both clients below:
```
which ucsc-mcp
```
Claude Code — register the stdio server:
```
claude mcp add ucsc-cell-browser /path/to/ucsc-mcp
```
(replace /path/to/ucsc-mcp with the absolute path that which ucsc-mcp printed — e.g. /Users/you/.local/bin/ucsc-mcp).

Claude Desktop — add to claude_desktop_config.json:

{
  "mcpServers": {
    "ucsc-cell-browser": {
      "command": "/path/to/ucsc-mcp"
    }
  }
}

Fully quit and relaunch Claude Desktop after editing.

This is a stdio server that Claude launches itself. Running ucsc-mcp in a terminal only verifies that it boots — press Ctrl-C afterwards and do not leave it running.

What it does

Parses the UCSC Cell Browser’s public dataset index (https://cells.ucsc.edu/dataset.json) and exposes it as queryable tools:

search_datasets — keyword search across the collection
get_dataset_details — full metadata for one dataset
list_organisms — organisms represented
list_body_parts — tissues / body parts represented
list_diseases — disease annotations represented
list_projects — parent projects / consortia
refresh_data — re-pull the cached dataset index

Primary use cases: locating a public scRNA-seq dataset for a tissue or disease, scoping what single-cell data already exists before generating new data, assembling dataset shortlists for reanalysis.

Notes

This is a discovery-and-metadata layer, not a data-access layer: it tells you which datasets exist and what they cover, but does not download expression matrices. Pair it with cellxgene-census or NCBI GEO to fetch the underlying counts, and with Scanpy to analyse them.

Dependencies are light (httpx, mcp). No API key or account is needed because the backing index is public.

The repository is small (1 star) and has not been updated since 2025-07-28; the license discrepancy between the LICENSE file (BSD-3-Clause) and pyproject.toml (MIT) is unresolved upstream. Both are permissive, but confirm with the maintainers before redistributing.

The server is one of the reference implementations described in the MCPmed paper (Briefings in Bioinformatics, 2026), alongside the already-catalogued NCBI GEO and STRING servers from the same group.

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.