Jevry splits browser automation into three separate jobs: a language model plans what to do, Jev converts that plan into a structured, typed decision with a guard around it, and Chromium actually performs the action.
Every action is observable rather than a black box: the decision, the guard, and the result of carrying it out are all visible as the task runs.
It ships as a real desktop browser you install and use, not a library you wire into your own app, with a short introduction video showing it working end to end.
What you get
- Full source code under the MIT license
- The Jevry desktop browser, its plan-decide-act pipeline and guard rails
CodeRabbit participant
Jev
JavaScript
Chrome
TypeScript
WebMCP
WXT