Skip to content

Model Context Protocol (MCP) Quickstart โ€‹

Connect your local AI agents (Claude Desktop, Cursor, AutoGPT) directly to live exchange order flow via the official BitCLI MCP Server.


1. Agentic Economy (HTTP 402) โ€‹

BitCLI implements a Machine-to-Machine (M2M) billing model using the HTTP 402 Payment Required standard. Your AI agent is charged in micro-cents (via Lightning Network or TON) per individual API call. You don't need a monthly subscriptionโ€”your agent pays only for the exact data it consumes during its trading loop.


2. Core Tools Exposed โ€‹

get_market_intel โ€‹

Returns institutional market state (Aggregated OI, RVOL, CVD, Funding) for any ticker:

json
{
  "command": "get_market_intel",
  "coin": "BTC"
}

get_top_radars โ€‹

Allows your agent to scan the entire market for anomalies without downloading raw candle data.

json
{
  "command": "get_top_radars",
  "category": "squeeze" // or "whales", "heat", "volatility"
}

create_alert โ€‹

Allows your agent to programmatically set real-time triggers on the C++ core:

json
{
  "command": "btc 95k"
}

3. System Prompt Recipe โ€‹

Paste this into your trading agent or Cursor system prompt:

text
You are an institutional crypto quantitative trader.
Before suggesting any perp trade setup, ALWAYS call the `get_market_intel` tool to check the asset's RVOL and CVD slope.
To find daily opportunities, query the `get_top_radars` tool with category "whales" or "squeeze".
Never calculate position size manually; use the `calculate_position_size` tool.