Hey Product Hunt, I'm Zach, founder of Keel. I've been using AI agents to build systematic strategies on Hyperliquid for over a year, and slowly built up a reliable system to represent, test and deploy them. Typically there are two options: 1) Put the LLM in the execution loop, so it makes the decisions and trades. That's slow, inconsistent and hard to audit. 2) Have the agent write all the code and the backtest. It makes subtle mistakes with time series data and timeframes, rarely reuses what worked, and it's hard to be confident the result is right, or to check it quickly yourself. With Keel we built a small typed language for strategies. Your agent builds a strategy from about 250 tested components, as a graph of steps that carries types and clocks. Drafts are validated before anything runs, so the agent gets a fast feedback loop and a lot of common errors never make it to a backtest. Each valid strategy is fingerprinted, and the exact same artifact runs in the backtest, in paper trading and live. Strategies are also much easier for you to verify, since they're made of self-describing components and mostly linear to read. You can try it from the Claude or ChatGPT directory, or in Claude Code, Cursor or Codex. Paper trade on live prices first, then go live on your own account from the Keel app. We're starting with Hyperliquid and adding more venues later. If you share trading ideas with an audience, you can also publish strategies your followers can run, and earn when they trade live. Would love to hear the first strategy you'd have your agent build, and what you'd want it to do that it can't yet. I'll be around all day answering questions.

Keel

Agent-native systematic trading, from backtest to live

Agent-native systematic trading, from backtest to live