Skip to content

Claude Desktop Setup

Connect the COT Bias MCP Server directly into Anthropic's Claude Desktop to give Claude real-time institutional CFTC positioning intelligence, AI market regimes, and macro attribution.

⚡ Setup in 3 Minutes

1. Requirements

  • Claude Desktop installed on macOS or Windows.
  • Python 3.10+ and uv installed (curl -LsSf https://astral.sh/uv/install.sh | sh).
  • A Quant Tier API key from COT Bias Terminal.

2. Configure Claude Desktop

Open your Claude Desktop configuration file:

code ~/Library/Application\ Support/Claude/claude_desktop_config.json

Open %APPDATA%\Claude\claude_desktop_config.json in your favorite editor.

Add the cotbias server under mcpServers:

{
  "mcpServers": {
    "cotbias": {
      "command": "uvx",
      "args": [
        "cotbias[mcp]",
        "cotbias-mcp"
      ],
      "env": {
        "COT_API_KEY": "cot_live_YOUR_ACTUAL_API_KEY"
      }
    }
  }
}

Replace cot_live_YOUR_ACTUAL_API_KEY with your actual Quant API key.

3. Restart Claude Desktop

Completely quit and restart Claude Desktop.

Look for the hammer icon (🔨) at the bottom-right of the prompt input window. Clicking it should list the 5 COT Bias tools: - get_market_snapshot - get_ai_inference - get_timeseries - scan_markets - list_markets

💬 Example Prompts to Try

Once connected, you can ask Claude natural language questions. Claude will automatically invoke the appropriate tool:

"Give me a weekly macro briefing on institutional COT positioning across Equities, Commodities, and Rates."

Claude will call scan_markets and synthesize extreme positioning and squeeze risks.

"What is the current institutional positioning regime in Gold (GC)? Are commercial hedgers accumulating or distributing?"

Claude will call get_market_snapshot or get_ai_inference for Gold.

"Scan the futures universe for contracts with High Squeeze Risk and tell me which markets have speculative shorts trapped."

Claude will call scan_markets(crowdedness_risk="High Squeeze Risk").

🛡 Quota Protection

Claude conversations often involve multi-turn questions about the same asset.

The COT Bias MCP server includes an automatic in-memory TTL cache (5 minutes for regimes/screeners, 1 hour for timeseries), ensuring that multiple follow-up questions do not drain your 2,000 requests/day API quota.