MCP Prediction Markets
# MCP Prediction Markets
Live prediction market intelligence for AI agents via the [Model Context Protocol](https://modelcontextprotocol.io/).
## Tools
| Tool | Description |
|------|-------------|
| `search_prediction_markets` | Search markets by keyword across Polymarket & Kalshi |
| `get_trending_markets` | Top markets by volume and activity |
| `get_market_odds` | Current odds and pricing for a specific market |
| `find_arbitrage` | Cross-platform price discrepancies (Kalshi vs Polymarket) |
| `get_category_markets` | Browse markets by category (crypto, politics, economics) |
All data from free public APIs (Polymarket, Kalshi). No API keys required.
## Install
```bash
pip install mcp-prediction-markets
```
## Usage with Claude Desktop
Add to your Claude Desktop config (`~/Library/Application Support/Claude/claude_desktop_config.json`):
```json
{
"mcpServers": {
"prediction-markets": {
"command": "mcp-prediction-markets"
}
}
}
```
Then ask Claude things like:
- "Search for Bitcoin prediction markets"
- "What are the trending markets on Polymarket?"
- "Find arbitrage opportunities between Kalshi and Polymarket"
- "What are the odds on the Fed cutting rates?"
## License
MIT
TDQS
Scored across 5 tools
Each tool has a distinct purpose: find_arbitrage detects cross-platform arbitrage, get_category_markets browses by category, get_market_odds fetches specific odds, get_trending_markets shows trending markets, and search_prediction_markets searches by keyword. No overlap.
All tool names follow a consistent verb_noun pattern in snake_case. Most use 'get_' prefix, while 'find_arbitrage' uses 'find_' which is still a clear verb_noun construction. Naming is predictable and uniform.
Five tools cover the essential information retrieval needs of a prediction markets server: discovering, searching, browsing, and checking odds, plus a unique arbitrage detection tool. The count is well-scoped and not excessive.
The tool set covers major read operations: search, browse, trending, odds, and arbitrage. However, it lacks tools for historical data, market analysis, or user portfolio management, which are reasonable extensions for a more complete prediction markets server.