Cellxgene Census (Claude Skill)

Claude skill that drives the CZ CELLxGENE Discover Census — a TileDB-SOMA-backed atlas of standardized human and mouse single-cell RNA-seq data — for population-scale subsetting, metadata filtering, and AnnData export.

   
Type Claude Skill
Supplier K-Dense Inc. (community OSS)
Availability GA — actively maintained 2025–2026
Pricing Free / OSS skill (MIT collection); the CELLxGENE Census itself is free / CC-BY
Capabilities Read-only — queries the public census; Claude executes Python locally

How to install

  • Also packaged in the SciAgent-Skills collection (jaechang-hits (community OSS, CC BY 4.0)): clone jaechang-hits/SciAgent-Skills and run /plugin install sciagent-skills in Claude Code (or copy skills/genomics-bioinformatics/single-cell/cellxgene-census into ~/.claude/skills/).
  • Claude Code / Claude.ai — Skills CLI (recommended): 
    npx skills add K-Dense-AI/scientific-agent-skills

    Installs the K-Dense collection; enable the cellxgene-census skill when prompted (also works in Cursor/Codex via the Agent Skills spec; requires Node ≥ 18).

  • Claude Code / Claude Desktop — manual clone: 
    git clone https://github.com/K-Dense-AI/scientific-agent-skills
cp -r scientific-agent-skills/skills/cellxgene-census ~/.claude/skills/
pip install cellxgene-census

Project-scoped alternative: copy into .claude/skills/ instead of ~/.claude/skills/.

What it does

SKILL.md with recipes for:

  • Opening the public census via cellxgene_census.open_soma() with version pinning
  • Metadata filtering (tissue, cell type, disease, assay, donor) before download to minimise transfer
  • Pulling cell- or gene-axis subsets directly into AnnData / Scanpy
  • Cross-dataset cell-type or disease comparisons across 1,000+ studies
  • Local caching and batched reads for memory-constrained workflows
  • Building reference atlases for downstream scVI / scANVI integration

Primary use cases: Reference-atlas construction for scVI / scANVI integration, cell-type-of-interest meta-analyses across public datasets, disease-vs-healthy expression comparisons, harmonised cohort assembly for downstream DE.

Notes

Pairs naturally with the catalog’s scanpy, anndata, scvi-tools, and scvelo entries for end-to-end single-cell workflows. Census versions are immutable snapshots; record the version you opened to keep analyses reproducible. Skill is documentation plus Python recipes — Claude calls cellxgene-census locally via Bash/Python, not as an MCP tool.

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.