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

1. Run the config generator:

   cd crow
   npm run desktop-config

2. Copy 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

3. 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-projects": {
      "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 automatically delivers behavioral context when Claude Desktop connects via MCP — memory protocols, session management, and transparency rules are active from the first message. No manual loading required.

For deeper guidance, Claude Desktop can use MCP prompts: session-start, crow-guide, research-guide, blog-guide, sharing-guide.

You can also manually load the full context:

"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.