We build the systems that make AI agents engineer like principals — done-state contracts, design law, and structure instead of vibes.
LLMs don't fail because they can't code. They fail because nobody told them what done means, what good looks like, and when to stop. Inerate exists to close that gap: every loop needs a base case —
if (done) return;
| Project | What it is |
|---|---|
| Atelier | The craftsman's framework for AI-built software: loop engineering with base cases, enforced clean-code law, anti-AI-slop design law, automatic model routing, SQLite knowledge graph, self-learning overlays. One install. |
/plugin marketplace add INERATE/atelier
/plugin install atelier
/atelier:bootstrap
- Done-states, not vibes — every task terminates against a written, checkable contract.
- Laws, not suggestions — 250 words per file enforced by hooks; one accent color enforced by review; no polling, anywhere, ever.
- Token economy — knowledge lives in queryable rows and graphs, not re-read documents.
- Premium by default — if it looks like a template, we delete it.
Every repo ships with a contribution guide, and every contribution passes the same gates our frameworks enforce on generated code. Start with an issue — we answer.
Founded and led by Piyush Sharma (owner) · MIT licensed · Built in the open