I've been building AI agent tools and noticed a gap: there's no easy way for AI agents to access structured crypto trading data via MCP (Model Context Protocol). So I built one.
What is BotIndex?
BotIndex is an MCP server that exposes 22 tools across crypto, sports, and agentic commerce — designed for AI agents to consume directly.
The most popular endpoints so far are the Hyperliquid tools:
Hyperliquid Tools
| Tool | Description | Price |
|---|---|---|
botindex_hl_funding_arb |
Cross-exchange funding rate arbitrage (HL vs Binance/Bybit) | $0.05 |
botindex_hl_liquidation_heatmap |
Liquidation cluster heatmap by price level | $0.05 |
botindex_hl_correlation_matrix |
Perpetual correlation matrix | $0.05 |
botindex_hl_coin_analytics |
Deep analytics per coin | $0.05 |
Other Tools
- Sports: Live odds, line movements, prop bets, DFS optimizer, arb scanner
- Crypto: Token universe, graduation signals
- Solana: Metaplex Genesis launches
- Zora: Trending coins, creator scores, attention momentum
- Commerce: Agentic commerce protocol comparison (ACP, UCP, x402)
Quick Start
npx botindex-mcp-server
That's it. No API key needed for the free tier (5 requests/hour).
Claude Desktop Config
{
"mcpServers": {
"botindex": {
"command": "npx",
"args": ["-y", "botindex-mcp-server"]
}
}
}
Architecture
-
Transport: Stdio (default) + Streamable HTTP (
--httpflag for Smithery) - Backend: Express on Fly.io with file-based API key storage
- Payments: Stripe for API keys, x402 for per-request crypto payments
- Trust Layer: Every response signed with Ed25519 (Agent Action Receipt spec)
What Makes This Different
- Agent-first: No human UI. Every endpoint returns structured JSON optimized for LLM consumption.
- Cryptographic receipts: Every API response includes an Ed25519-signed receipt proving the data came from BotIndex. Published as an open RFC.
- Dual payment rails: Traditional API keys via Stripe AND x402 crypto micropayments. Pay however your agent prefers.
- Free discovery: The catalog endpoint is always free — agents can explore before committing.
Links
- Smithery: smithery.ai/servers/cyberweasel777/botindex-mcp-server
- npm: npmjs.com/package/botindex-mcp-server
- GitHub: github.com/Cyberweasel777/botindex-mcp
- AAR Spec: github.com/Cyberweasel777/agent-action-receipt-spec
Would love feedback from anyone building with MCP or agent tooling. What data sources would be most useful for your agents?
Top comments (1)
Nice MCP server — the tool descriptions and parameter docs are where the prompt engineering lives for these. Clear, constrained tool descriptions mean agents call the right tool with the right args on the first try. The quality of those descriptions matters as much as the underlying implementation. flompt helps with structuring those prompt-level descriptions too. flompt.dev / github.com/Nyrok/flompt