Quick Start

May 5, 2026 · View on GitHub

This page is for your first successful analysis after installation.


Your First Analysis

Open your MCP client and type:

Load /absolute/path/to/spatial_data.h5ad and show me the tissue structure

If you configured ChatSpatial through Docker, use the container path from your mount instead:

Load /data/spatial_data.h5ad and show me the tissue structure

For example, -v /Users/alice/spatial-data:/data:ro makes host files under /Users/alice/spatial-data visible as /data inside ChatSpatial. For the full Docker path model, see Docker / GHCR.

Then continue with:

Normalize and cluster the data
Identify spatial domains with SpaGCN

Common First Prompts

TaskSay This
Load dataLoad my Visium data from /absolute/path/to/data
PreprocessNormalize and cluster the data
Find domainsIdentify spatial domains with SpaGCN
Annotate cellsAnnotate cell types using the reference
DeconvolveEstimate cell type proportions
Find spatial genesFind spatially variable genes
VisualizeShow expression of CD3D on the tissue

Sample Data

Try these public test files from the sample-data release:

The sample-data release is versioned separately from the ChatSpatial package and Docker image.


What Success Looks Like

After loading:

Dataset loaded successfully
- ID: spatial_data_abc123
- 3000 spots, 18000 genes
- Platform: Visium

After preprocessing:

Preprocessing complete
- Filtered to 2800 spots, 2000 HVGs
- Computed 30 PCs, UMAP
- Found 8 clusters (Leiden)

Visualizations appear directly in the chat or client UI.


Before You Troubleshoot

Check these basics first:

  • Use an absolute data path, not ~/... or ./...
  • If using Docker, use the container path from your mount, such as /data/sample.h5ad; full Docker mount rules live in Docker / GHCR
  • Restart your client after MCP configuration changes
  • Run preprocessing before most downstream analyses

If that does not fix the problem, go to the Troubleshooting Guide.


Next Steps