About

About rightmodeler

Proof over guesswork, for the models your agents run on.

rightmodeler is an open-source tool for teams running multi-agent LLM systems. It answers one question with evidence: which model calls can move to a cheaper model without losing quality?

The mission is simple: no model decision on vibes. rightmodeler detects inefficient calls, proves the safe swaps on your own traces against what you already shipped, and applies the fix in your repo. A report and an edit, never a runtime gateway.

rightmodeler is an ELM-OS project. The skill is available today; rightmodeler agent, which ships swaps as pull requests, and Crucible, the analytics and optimization suite, are built on the same proof loop and coming next.

Find us

FAQ

What is rightmodeler?
An open-source tool for teams running multi-agent LLM systems. It proves, on your own traces, which model calls can move to a cheaper model without losing quality, then applies the safe swaps in your repo.
Is it open source?
Yes. The rightmodeler skill is on GitHub and free to run on your own traces today. rightmodeler agent, which opens evidence-backed swap PRs, and Crucible, the analytics and optimization suite, are the products being built on top of it.
How is it different from observability or a gateway?
Observability shows you problems; a runtime gateway hijacks live traffic. rightmodeler proves the fix on your own traces and applies it in your repo: detect, prove, fix. Nothing runs in your request path.