Stage‑aware Codex orchestrator that routes AI stages with evidence
Codex Stage Orchestrator lets developers assign distinct AI models to specific stages of a coding task—planning, design, coding, review, etc.—while automatically recording which model actually ran. It runs as a lightweight Node.js MCP server, exposing a concise control card inside Codex that shows current stage, requested route, and evidence. The tool is ideal for teams that need auditability and consistency across AI‑assisted workflows, and it outperforms ad‑hoc routing scripts by providing built‑in enforcement and a single source of truth.
View on GitHub →slegarraga/codex-stage-orchestrator