Agents spend expensive model time on choices that are not really thinking: which model should answer this turn, which of hundreds of skills to load, which memories matter, what survives a summary. This package hands those choices to Jev.
It covers model routing, memory filtering, compaction, skill selection, and browser and computer use, and comes with a dashboard showing each routing decision live. You can switch Jev on, off, or into a shadow mode that only watches.
Includes a private flow for entering your TypeSafe key.
What you get
- Full Python source code, MIT licensed
- Skills, router dashboard and installer
- Evals, docs and tests






