My friend Anton mentioned this to me. I haven’t used it yet, but it looks interesting.
Ralphex is a CLI that runs AI coding agents like Claude Code and Codex through a plan you write, one task at a time. Each task gets a fresh session, so the agent’s context doesn’t fill up with earlier work. After each one, Ralphex runs your tests and linters and commits.
Once every task is done, the review happens in rounds. First, five agents review the work in parallel, covering quality, implementation, testing, simplification and documentation, and Claude fixes what they find. Then GPT-5 gives a second opinion through Codex, going round until Codex reports no issues (it’s skipped if you don’t have Codex installed). Last, the quality and implementation agents do a final pass for anything critical. There’s also a real-time web dashboard for watching it run.
flowchart TD
subgraph each [For each task in your plan]
direction LR
session[Fresh session] --> checks[Tests and linters] --> commit[Commit]
end
subgraph after [Once every task is done]
direction LR
review[Five reviewers] --> codex[GPT-5 via Codex] --> final[Final review]
end
each --> after
The tagline is “Write a plan. Walk away.” With a commit per task and the review results to read through, you can see what it did while you were gone.