MCP Server for AI Agents

prediction.click provides an MCP (Model Context Protocol) server that lets AI agents access real-time fair values, market data, and trading opportunities programmatically.

Repository: github.com/clober-dex/prediction-click-mcp-server

What Can Agents Do?

With the MCP server, an AI agent can:

• Scan all 24 markets for the best trading opportunities

• Get real-time fair values and market prices

• Read Polymarket orderbooks and find profitable fills

• Monitor spot prices and volatility from Binance

• Check system health status

Setup

Claude Desktop

Add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "prediction-click": {
      "command": "npx",
      "args": ["-y", "prediction-click-mcp-server"]
    }
  }
}

Claude Code

claude mcp add prediction-click -- npx -y prediction-click-mcp-server

Cursor

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "prediction-click": {
      "command": "npx",
      "args": ["-y", "prediction-click-mcp-server"]
    }
  }
}

Run from Source

git clone https://github.com/clober-dex/prediction-click-mcp-server.git
cd prediction-click-mcp-server
npm install
npm run build
npm start

For development with hot reload:

npm run dev

Available Tools

scan_opportunities

The best starting point. Scans all 6 assets across all 4 timeframes and returns the most profitable opportunities ranked by edge.

Example prompt: "Scan all markets and tell me the best trading opportunities right now"

Example response:

{
  "scanned_at": "2026-03-20T12:00:00.000Z",
  "total_opportunities": 5,
  "opportunities": [
    {
      "asset": "BTC",
      "timeframe": "5m",
      "side": "UP",
      "edge": 4.21,
      "fair_value": 53.21,
      "market_price": 49.00
    }
  ]
}

get_markets

Returns the full grid of all market data — fair values and market prices for every asset and timeframe at once.

Example prompt: "Show me all the current markets"

get_fair_value

Get detailed fair value data for a specific asset and timeframe. Includes model parameters, volatility (sigma), and Polymarket token IDs needed for trading.

Parameters:

Name	Values	Description
asset	BTC, ETH, SOL, XRP, DOGE, BNB	Crypto asset
timeframe	5m, 15m, 1h, 4h	Market timeframe

Example prompt: "What's the fair value for BTC 5m right now?"

get_orderbook

Get the Polymarket CLOB orderbook (bids and asks) for a specific market and side.

Parameters:

Name	Values	Description
asset	BTC, ETH, SOL, XRP, DOGE, BNB	Crypto asset
timeframe	5m, 15m, 1h, 4h	Market timeframe
side	yes, no	yes = UP token, no = DOWN token (default: yes)

Example prompt: "Show me the orderbook for ETH 1h UP side"

get_opportunity

Find profitable fills for a specific market by walking the orderbook. Returns fills where the market price is below the fair value (positive edge), including expected profit after Polymarket fees.

Parameters:

Name	Values	Description
asset	BTC, ETH, SOL, XRP, DOGE, BNB	Crypto asset
timeframe	5m, 15m, 1h, 4h	Market timeframe

Example prompt: "Are there any profitable fills for SOL 15m?"

get_spot_price

Get the current real-time spot price from Binance, plus recent 1-minute candle history (~8 hours).

Parameters:

Name	Values	Description
asset	BTC, ETH, SOL, XRP, DOGE, BNB	Crypto asset

Example prompt: "What's the current BTC price?"

get_history

Get historical fair value and market price data at 1-second granularity. Useful for analyzing how fair value and market price have diverged over time.

Parameters:

Name	Values	Description
asset	BTC, ETH, SOL, XRP, DOGE, BNB	Crypto asset
timeframe	5m, 15m, 1h, 4h	Market timeframe

Example prompt: "Show me the recent history for DOGE 4h"

get_health

Check the health status of all prediction.click backend services (price daemon, market daemon, pricing engine).

Example prompt: "Is prediction.click running normally?"

Typical Agent Workflow

A typical trading analysis workflow looks like this:

1. scan_opportunities
   → Find the best edges across all markets

2. get_opportunity (for top picks)
   → See exact fills and expected profit after fees

3. get_orderbook (check liquidity)
   → Verify there's enough depth to fill your order

4. get_fair_value (for context)
   → Check volatility, model params, and token IDs

Rate Limits

The prediction.click API allows 600 requests per minute per IP address. Exceeding this limit returns HTTP 429 with a Retry-After header.

Important Notes

• All prices are probabilities in the 0–1 range (displayed as percentages)

• Edge is in percentage points (e.g., 3.2 means a 3.2% edge)

• Markets rotate frequently — a 5m market expires every 5 minutes

• If stale: true appears in the data, the pricing engine has outdated data — skip those markets

• The MCP server is read-only — it does not place trades. Use the prediction.click web UI to execute trades