Claude Desktop
Connect Crow to Claude Desktop using local stdio transport. This runs the MCP servers directly on your machine — no cloud deployment needed.
Prerequisites
- Claude Desktop installed
- Crow cloned and set up locally (Desktop Setup Guide)
Setup Steps
Run the config generator:
bashcd crow npm run desktop-configCopy the output JSON into your Claude Desktop config file:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json - Linux:
~/.config/Claude/claude_desktop_config.json
- macOS:
Restart Claude Desktop
The config will look something like:
{
"mcpServers": {
"crow-memory": {
"command": "node",
"args": ["/path/to/crow/servers/memory/index.js"],
"env": {
"CROW_DB_PATH": "/path/to/crow/data/crow.db"
}
},
"crow-research": {
"command": "node",
"args": ["/path/to/crow/servers/research/index.js"],
"env": {
"CROW_DB_PATH": "/path/to/crow/data/crow.db"
}
}
}
}Transport
- Type: stdio (direct process)
- Auth: None needed (local process)
Verification
After restarting Claude Desktop, look for the MCP server icons (hammer icon) in the input area. Click them to see available tools.
Try: "Store a memory that Crow Desktop is connected."
Cross-Platform Context
Crow's shared behavioral context (crow.md) is available through the crow_get_context tool or crow://context resource. Load it at session start:
"Load your crow.md context"
Memories stored in Claude Desktop are shared with all other connected platforms (Claude Web, ChatGPT, Gemini, etc.) when using the same database. See the Cross-Platform Guide.
Adding External Integrations
To use external integrations (GitHub, Slack, etc.) with Claude Desktop, add them directly to the Desktop config. The .mcp.json file in the Crow repo has all the configurations — merge them into your Desktop config file.
You'll need to replace ${VAR_NAME} references with actual values or set them as environment variables.