Claude Code (CLI)

Connect Crow to Claude Code, Anthropic's CLI tool for using Claude in the terminal.

Option A: Local (stdio)

Best for development — runs Crow servers directly on your machine.

Setup Steps

1. Clone and set up Crow locally:

   git clone https://github.com/kh0pper/crow.git
   cd crow
   npm run setup

2. Add to your project's .mcp.json (per-project) or ~/.claude/mcp.json (global):

   {
     "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"
         }
       }
     }
   }

3. Restart Claude Code — it will automatically detect the MCP servers.

Transport

• Type: stdio
• Auth: None (local process)

Option B: Remote (HTTP)

Connect to a deployed Crow gateway for access to the full platform including external integrations.

Setup Steps

1. Deploy Crow (Cloud Deploy Guide)

2. Add to .mcp.json:

   {
     "mcpServers": {
       "crow-memory": {
         "type": "url",
         "url": "https://your-crow-server/memory/mcp"
       },
       "crow-projects": {
         "type": "url",
         "url": "https://your-crow-server/projects/mcp"
       },
       "crow-tools": {
         "type": "url",
         "url": "https://your-crow-server/tools/mcp"
       }
     }
   }

3. On first use, Claude Code will open the OAuth flow in your browser to authorize.

Transport

• Type: Streamable HTTP
• Auth: OAuth 2.1 (automatic)

Verification

Start Claude Code and try:

> Store a memory that Claude Code is connected to Crow

Check it worked:

> What do you remember?

Cross-Platform Context

Crow automatically delivers behavioral context when Claude Code connects via MCP — memory protocols, session management, and transparency rules are active from the first message.

Claude Code users also have access to skill files in skills/ and the project-level CLAUDE.md, which provide additional platform-specific guidance beyond what MCP instructions deliver.

For detailed guidance via MCP, use prompts: session-start, crow-guide, research-guide, blog-guide, sharing-guide. Or use crow_get_context with platform: "claude". Memories stored via Claude Code are shared with all other connected platforms. See the Cross-Platform Guide.

Already set up on claude.ai?

If you've added Crow as a Custom Integration on claude.ai, those MCP servers are also available in Claude Code CLI without additional configuration — Anthropic's ecosystem shares connector settings across Claude products. No need to duplicate setup in .mcp.json. This sharing behavior is specific to Claude; other platforms (ChatGPT, Gemini) manage their MCP connections independently.

Tips

• Use the project-level .mcp.json to share Crow config with your team
• Use ~/.claude/mcp.json for global access across all projects
• The Crow repo itself includes a .mcp.json with all MCP servers pre-configured