About

About OpenAdapt

OpenAdapt compiles demonstrated GUI workflows into deterministic, locally executable programs. Healthy runs make no model calls. When the UI drifts, retained evidence may re-resolve the target; optional AI may propose a repair, and configured verification can halt instead. Repairs are governed, reviewable diffs rather than an unconstrained promise of autonomous repair.

Who's behind it

OpenAdapt is built by MLDSAI Inc., founded in Toronto in 2023. The founder is Richard Abrich, a machine learning engineer who has been building OpenAdapt in the open since the first commit. There's no stealth roadmap. The issues, the pull requests, and the mistakes are all public on GitHub.

Open source, actually

The canonical compiler and governed runtime live in openadapt-flow. The OpenAdapt launcher provides the stable install and unified CLI, while optional privacy and research packages remain separate surfaces. The engine and launcher are MIT-licensed: use them, fork them, and ship them without asking permission.

How to reach us

Email hello@openadapt.ai and a person will reply. For bugs and feature requests, open an issue on GitHub. For everything else, there's Discord.

Talk to us

Bring one repeated, high-friction workflow to a 30-minute call. We'll map the workflow, verification oracle, execution boundary, and fastest path to a production deployment.

Evaluate a workflow