I built Keel because writing one-off scripts with AI or putting LLMs in the execution loop is not reliable, slow and hard to audit. You end up with lots of subtle errors with time series data.
Instead in Keel you build a strategy from typed components, can validate it before it runs, and the same compiled strategy runs in backtests and live on Hyperliquid.
Instead in Keel you build a strategy from typed components, can validate it before it runs, and the same compiled strategy runs in backtests and live on Hyperliquid.
We explain more how it works here https://usekeel.io/how-it-works
Still working on improving the kinds of strategies that can be built here and how well agents go from natural language to a strategy.