crypto-intel-mcp

Crypto market intelligence for any LLM, over the Model Context Protocol.

Gives any MCP client (Claude Desktop, Claude Code, agents) clean tools for crypto price, momentum, and what's trending — and blends them into a single scored verdict. Runs entirely on CoinGecko's free public API (no key), fails gracefully (a missing source returns "no data", never a fabricated value), and outputs Markdown by default or JSON on demand.

# Crypto Intelligence — BTC

**Signal:** Lean positive  `[........|##......]` +22/100  · confidence 80%

Building momentum → LEAN LONG. BTC momentum: 24h +1.9% · 7d +4.4% · 30d -17.0%.
- In CoinGecko trending (elevated attention).
- Total crypto market cap +1.8% 24h, BTC dominance 57%.

Tools

Every tool returns Markdown (human-readable, default) or JSON (response_format="json") for programmatic use.

Live demo output (2026-06-15)

crypto_get_quote bitcoin:
  BTC $66,504.00  +3.4% 24h  · rank #1  · -47% from ATH

crypto_momentum bitcoin:
  Lean positive  [........|##......] +21/100  · confidence 80%
  BTC momentum: 24h +3.4% · 7d +4.9% · 30d -14.8%

crypto_trending:
  ZEC, PENGU, SIREN, TAO, NEAR, BTC, HYPE, GRAM, SUI, H

crypto_analyze bitcoin:
  Lean positive  [........|###.....] +35/100  · confidence 80%
  Building momentum → LEAN LONG. In CoinGecko trending (elevated attention).
  Total crypto market cap +3.6% 24h, BTC dominance 57%.

All values are live from CoinGecko's free public API — no key, no fabrication.

Quick start

git clone https://github.com/cstamigo-droid/crypto-intel-mcp crypto-intel-mcp
cd crypto-intel-mcp
python -m venv .venv && .venv\Scripts\activate     # Windows
pip install -r requirements.txt

python -m crypto_intel_mcp     # starts the MCP server over stdio (no .env needed)

Smoke test (hits the live CoinGecko API and prints each signal):

python tests/test_smoke.py bitcoin

Use it in Claude Desktop

Add this to claude_desktop_config.json (%APPDATA%\Claude\ on Windows, ~/Library/Application Support/Claude/ on macOS), then restart Claude Desktop:

{
  "mcpServers": {
    "crypto-intel-mcp": {
      "command": "python",
      "args": ["-m", "crypto_intel_mcp"],
      "cwd": "C:/path/to/crypto-intel-mcp"
    }
  }
}

Why it's built this way

• Uniform result contract. Every source returns the same Result shape, so an LLM can reason across tools instead of parsing N formats.
• Graceful degradation. A source with no data returns "no data", not a fake value. Missing data never invents an answer.
• Resilient + cached. A short per-source TTL cache avoids hammering rate-limited endpoints when an agent calls several tools in one turn.

Disclaimer

For research and educational use only. Data comes from third-party sources and may be delayed or incomplete.

License

MIT