PolyTerminal autonomous trader

A developer has published PolyTerminal, an open-source bot that scans Polymarket, asks ChatGPT for an odds view, sizes bets with the Kelly criterion, and can place trades automatically. (github.com) The code is live in a public GitHub repository, where the README describes PolyTerminal as an “autonomous AI trading agent” for Polymarket and lists analysis, decision, and execution modules. The repo says it needs Python 3.8+, an OpenAI key, a SerpAPI key, a Polygon wallet with MATIC for gas, and USDC for trading. (github.com) In the repository’s workflow, the bot compares its own probability estimate with the market price, looks for a gap it treats as an edge, and then calculates stake size with the Kelly criterion before sending orders. The README says execution runs through Polymarket’s Agents software development kit, with risk controls such as bankroll settings and maximum bet percentages in the config. (github.com) Prediction markets turn event forecasts into tradeable contracts, with prices between $0 and $1 that function like probabilities. Polymarket’s developer docs say builders can fetch markets, place orders, and redeem positions through official application programming interfaces and software development kits in Python, TypeScript, and Rust. (docs.polymarket.com) That plumbing has made it easier for independent developers to build automated traders on top of Polymarket instead of clicking through the website by hand. Polymarket also maintains a public “agents” repository that it describes as a framework and set of utilities for building artificial intelligence agents for prediction markets. (github.com) The Kelly criterion is a bankroll formula used to decide how large a bet should be when a trader thinks the odds are in their favor. PolyTerminal’s README says the bot uses Kelly-based sizing for “optimal bankroll management,” which means the system is not just picking markets but also deciding how much capital to risk on each one. (github.com) The release lands amid a wider burst of Polymarket automation projects. Recent public repositories and demos describe bots that scan hundreds of markets, estimate probabilities with large language models, and execute trades through Polymarket’s central limit order book, or live order-matching system. (github.com 1) (github.com 2) That does not settle whether these systems make money. PolyTerminal’s public materials describe features and setup steps, but they do not publish audited returns, live performance records, or independent evidence that ChatGPT-based forecasts beat Polymarket prices after fees and errors. (github.com) For now, PolyTerminal is a clear example of where this corner of crypto is heading: prediction markets as machine-readable feeds, language models as forecasters, and bet sizing turned into code. The practical test will be whether automated agents can keep finding mispriced contracts once more traders start running similar software. (docs.polymarket.com) (github.com)