AI agents
AI trading agents for prediction markets
LLM-driven agents that research and trade Polymarket and Kalshi markets, with what each actually automates and how well it holds up.
Checked September 8, 2026 · every price read off the live site
Every ai agents bot
3 reviewed · alphabetical-
05CloddsBot Open-source Claude-based trading agent across ten venues with Kelly sizing and circuit breakers, pushed daily, no profit claims.
-
11nothing-ever-happens Buys NO on every non-sports market. Self-described as mostly a meme, but it operationalizes a real tendency: roughly 70% of markets resolve NO.
-
Alphapoly (polymarket-alpha-bot) Finding logically related Polymarket markets with an LLM and entering the covering pair when it costs less than it pays.
| Bot | Venues | Pricing | Type |
|---|---|---|---|
| CloddsBot | Polymarket, Kalshi, Betfair, Smarkets, Drift, Opinion.xyz, Predict.fun, Manifold, Metaculus, PredictIt | Open source (MIT). | bot |
| nothing-ever-happens | Polymarket | Open source (CC0-1.0) | bot |
| Alphapoly (polymarket-alpha-bot) | Polymarket | Open source (Apache-2.0). | web app |
How these bots work
AI bots use a language model somewhere in the loop: to research a market and produce a probability, to find logical links between markets, or to act as a conversational agent that reads books and places orders. The strategy is only as good as the model's calibration, and none of the entries here publishes live calibration.
Two shapes appear. Agents such as CloddsBot put the model in front of many venues and let you talk to it. Screens such as Alphapoly use the model for one job (implication extraction) and leave the trade decision to price math and a person.
nothing-ever-happens sits here as the counterexample: a one-rule bot whose entire model is a prior about how markets resolve.
How to choose
- Ask what the model decides. Probability estimates, market matching, or order placement are different risk levels.
- Dry-run defaults and approval gates for anything that spends money or runs shell commands.
- Provider flexibility and cost: Anthropic-only versus 8 providers, per-report credit pricing, tokens per pipeline run.
- A decision ledger. Bots that log the model's reasoning per trade can be audited after a loss.
- Scope discipline. A README that also covers token launches is a README you will have to read all of.
Questions about ai agents bots
What is the best AI trading bot for Polymarket?
CloddsBot has the most stars (901) and the widest venue list. Alphapoly is narrower and current on Polymarket V2. Neither publishes returns.
Is there a Claude-based prediction market bot?
CloddsBot is built on Claude with an MCP server; kalshi-trading-bot-cli and DRADIS accept Anthropic keys among other providers.
Do AI prediction market bots work?
The READMEs here make no profit claims. kalshi-trading-bot-cli ships a backtest command that scores its model's Brier score against the market, which is the right question to ask any of them.
Bots die in execution. Terminals are execution.
Wanted a bot for speed and cross-venue vision? The free Kairos terminal trades Kalshi, Polymarket and Predict.fun from one book, no keys to manage.