conductor

Why we included this project

Most people I know who use coding agents have the same complaint: the output is often good, but it's also unpredictable, and it's hard to tell what the model will do on any given request. Conductor fixes that by adding a discipline layer on top of agents like Antigravity and Claude Code. Before the agent writes any code, it has to capture the project context, produce a spec and a plan, and only then implement against that plan. Features and bug fixes all go through the same Context -> Spec & Plan -> Implement lifecycle, so the repository becomes a single source of truth the agent refers back to instead of improvising from scratch each session. That's genuinely useful for teams that want changes they can review and reproduce, and for repos where several people or agents are touching the same code. You keep your existing agent; it just starts acting like a careful project manager.