A loop system for running several AI coding agents
A new setup lets Codex, Claude, Gemini, and similar run as multiple workers on the same . The main idea is not to give AI more freedom, but to surround it with strict, ordinary that real projects need. Each loop starts with fresh context, so the worker is less tied to an earlier conversation.
A human sets hard limits on the number of loops or minutes before the system stops. The setup can require checks such as pylint, ruff, pyright, Pydantic, and semgrep, and it can block the AI from changing certain files. Developers can add extra personal or team rules in preferences.py when existing tools cannot catch them.
Examples include limiting a file to no more than five public functions or banning unwanted loop patterns. Before an AI worker can mark the job as done, it must show proof from the repository, and the system can connect to git hooks for automatic checks.
Key points
- The setup can run Codex, Claude, Gemini, or other as repeated workers.
- Each loop starts with fresh context to reduce carryover from earlier .
- A human sets hard limits for time or number of iterations.
- Checks such as pylint, ruff, pyright, Pydantic, and semgrep can be required before work is accepted.
- preferences.py can hold personal or team coding rules that normal tools may not catch.