Server Deployment Guide
Deploy ChatSpatial on a server for team-wide access to spatial transcriptomics analysis.
MCP Server Deployment
ChatSpatial implements the Model Context Protocol (MCP) with two transport modes:
Local Mode (stdio)
For Claude Desktop and local clients:
# Install ChatSpatial
pip install -e ".[full]"
# Run in stdio mode (default)
python -m chatspatial
Remote Mode (SSE)
For remote access and HTTP-based clients:
# SSE mode with HTTP/WebSocket support
python -m chatspatial server --transport sse --port 8000 --host 0.0.0.0
Note: Use with mcp-remote proxy for Claude Desktop remote connections:
# Client-side proxy setup
npx mcp-remote http://your-server:8000/mcp/
Requirements
- Python 3.10-3.12 (MCP requires Python 3.10+)
- 64 GB RAM recommended for large datasets
- Fast SSD storage
- Network access for team
Configuration
Create config.yaml for server settings:
server:
host: 0.0.0.0
port: 8000
max_connections: 10
data:
cache_dir: /data/chatspatial_cache
max_dataset_size: 10GB
Security Notes
- Use VPN or internal network only
- Add authentication if exposing to internet
- Regular backups of analysis results
Getting Help
- GitHub Issues for deployment questions
- Check logs in
/var/log/chatspatial/
Next: Configuration Guide