Agentic change control is a way to govern software changes made by AI coding agents. It extends traditional version control with the operational facts agents need to work safely over time: intent, task ownership, findings, blast radius, thread context, proposal review, and release readiness.
flowchart TD A[nool task create] --> B[nool task pick] B --> C[nool thread create] C --> D[nool announce intent] D --> E[Agent implementation] E --> F[nool learn] F --> G[nool debug blast-radius] G --> H[nool propose] H --> I[nool solidify] I --> J[nool doctor]
The category exists because agents can modify many files quickly. They can also forget why a task was bounded, duplicate previous review issues, or merge changes that are textually clean but semantically incompatible. Agentic change control turns those risks into inspectable state.
The verified Nool cycle is deliberately explicit. Tasks are created with `nool task create --name ... --desc ... --solidify`, claimed with `nool task pick --id ... --solidify`, and finished with `nool task finish --id ... --solidify`. Threads are created with `nool thread create --name ...`, inspected with `nool thread show <name> --full`, and advanced with `nool thread status --name <name> --status review` or another supported status.
Implementation still happens in the normal tools: Claude Code, Codex, Cursor, a shell, or an editor. Nool governs the transition around that implementation using `nool announce intent`, `nool learn`, `nool findings`, `nool debug blast-radius`, `nool propose`, and `nool solidify`.
Good agentic change control does not make developers fill out ceremony for its own sake. It makes future sessions cheaper, reviews more cumulative, and releases more explainable.
The release side is part of the category too. `nool doctor --compact` tells the team whether semantic state, pending proposals, active threads, and filesystem hygiene are ready. `nool release <version> --solidify` records the release marker when the DAG is ready. That gives agent-authored work an auditable endpoint instead of an informal "looks good" moment.
That endpoint matters for audits and for future agents. A release marker says which semantic history was accepted as a coherent state, not just which commit happened to be latest.
It turns release into evidence.