COT Bias MCP Server (`cotbias-mcp`)¶

The COT Bias MCP Server implements Anthropic's Model Context Protocol (MCP) to connect institutional CFTC Commitments of Traders (COT) positioning regimes directly into AI agents (Claude Desktop, Cursor, Cline, LibreChat, and other MCP clients).

⚡ Quickstart¶

1. Requirements¶

A valid Quant Tier API key from COT Bias Terminal.
Python 3.10+ installed (uv or pipx recommended).

2. Client Configuration¶

Add the cotbias server to your client's configuration file (e.g. Claude Desktop or Cursor):

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

For client-specific guides: - Claude Desktop Setup - Cursor IDE Setup - Cline & Other MCP Clients

🛠 Available Tools¶

Tool	Parameters	Description
`get_market_snapshot`	`symbol` (e.g. `'ES'`, `'GC'`, `'CL'`, `'6E'`)	Recommended: Single call synthesizing the current week's AI regime, macro attribution, flow summary, and latest 4-week COT trend.
`get_ai_inference`	`symbol`	Institutional positioning regime, AI bias score, momentum, crowdedness risk, dumb money index, and macro rationale.
`get_timeseries`	`symbol`, `contract_type`, `lookback_weeks` (default 26, max 156), `report`	Historical weekly positioning data (commercial/non-commercial net, % of OI, 52-week COT index, 3-year Z-score, hedging pressure).
`scan_markets`	`ai_bias_score`, `crowdedness_risk`, `momentum`, `cot_positioning`	Multi-market screener across all tracked commodities, equities, FX, and rates.
`list_markets`	`sector` (optional: `'Equity'`, `'Commodity'`, `'FX'`, `'Rates'`)	List all tracked contracts with symbols, CFTC codes, and report mappings.

See Tools Reference for in-depth parameter schemas and response formats.

📦 Resources & Prompts¶

Resource cot://markets: JSON catalog of all tracked markets.
Prompt cot-macro-briefing: Guided prompt analyzing macro positioning shifts and institutional extremes across Equities, Commodities, and Rates.
Prompt cot-market-deep-dive: Deep-dive analysis for a specific ticker symbol.

🛡 Quota & Performance Guardrails¶

In-Memory TTL Caching: The server automatically caches tool responses (5 minutes for inference/screener, 1 hour for timeseries) to prevent AI chat sessions from rapidly consuming your daily 2,000 requests/day quota on weekly COT data. See Caching & Quotas.
Context-Optimized Payloads: Timeseries queries default to 26 weeks and project core analytical fields to keep prompt token consumption minimal.
Actionable Error Messages: If your key is invalid or your daily quota is exhausted, tools return clean structured errors with hints rather than crashing the agent. See Troubleshooting.

COT Bias MCP Server (cotbias-mcp)¶