---
type: tool
url: https://github.com/chaseai-yt/claudex-loop
category:
  - "[[Agent Methodology]]"
  - "[[Model Routing]]"
summary: "Cross-provider adversarial plan review (Claude ⇄ Codex)"
added: 2026-10-16
---

## Features
- Cross-provider adversarial plan review (Claude ⇄ Codex)
- Author never grades own work: builder ≠ inspector, fresh session
- SHA256-bound approvals (plan change → re-review required)
- Bounded rounds, append-only decision log (`PLAN-REVIEW-LOG.md`)
- Advisory model routing (`claudex-route`, human-gated)

## Notes