Overview
A minimal, open-source Python bot that turns Claude into a prediction market trader. It polls Kalshi for open markets every 5 minutes, asks Claude to evaluate each one, and places trades when it detects a meaningful edge between Claude’s probability estimate and the market price.
Runs on Kalshi’s free demo environment by default — no real money needed to try it.
Deploy in One Click
Or clone and run locally:
git clone https://github.com/rsquaredsolutions2026/kalshi-news-bot.git
cd kalshi-news-bot
cp .env.example .env # add your API keys
pip install -r requirements.txt
python bot.py
How It Works
Every 5 minutes:
1. POLL Fetch open markets from Kalshi
2. ANALYZE Ask Claude to evaluate each market
3. DETECT Compare Claude's estimate to market price
4. TRADE If edge > 5%, place a limit order
The bot skips sports markets (where Claude has no edge) and focuses on politics, economics, and current events where reasoning about public information matters.
Key Features
Claude-powered analysis — Sends each market to Claude with a structured prompt. Claude returns a probability estimate, confidence rating, and step-by-step reasoning.
Edge detection — Computes the confidence-weighted difference between Claude’s estimate and the market price. Only trades when the net edge exceeds a configurable threshold (default 5%).
Demo-first — Points to Kalshi’s free demo environment by default. Switch to production with one env var change.
Dry-run mode — Set DRY_RUN=true to log every decision without placing orders. Perfect for tuning.
~300 lines of Python — No frameworks, no async, no databases. Just a loop. Easy to read, easy to modify.
Requirements
- Python 3.10+
- Kalshi account with API keys (demo.kalshi.com for free testing)
- Anthropic API key (console.anthropic.com)
Links
- GitHub: rsquaredsolutions2026/kalshi-news-bot
- License: MIT