Vowline and Codex

Codex writes the code.
Vowline proves it works.

Codex is OpenAI’s coding agent: you give it a task and it changes your code. Vowline does not replace it: it gives Codex its tasks, runs its command line tool in a separate worktree, and checks every change on the exact commit before it counts.

CodexVowline
What it isA coding agent: it reads your repository and changes code for the task you give it.A control plane above coding agents: it plans, runs them and proves the result.
Where code is writtenWhere you run it, in your working copy or in its own environment.By the agent of your choice, in a separate git worktree per task.
Which modelsOpenAI’s models.Claude Code, Codex or Cursor, per task or per teammate.
What the product promisesNot its job: it works on the task in front of it.A map of roles, flows and steps, built from the code, with checkable criteria.
ProofTests run when you or the agent run them.Approved checks on the exact commit of every change; a promise drops back when its code changes.
TeamThe tasks you start yourself.A lead and specialists with areas of the code, chats and a backlog.
DecidingYou read what it changed and decide.You confirm criteria and accept changes; production deploys need your click.

They work together

Vowline runs the Codex command line tool with your own login, in its own sandbox and worktree. A teammate on Codex plans in the chat, takes a task with criteria you confirmed, and Vowline runs your approved checks on what it changed.

Vowline reads the 5-hour and weekly limits Codex reports and shows how much is left. When it runs out, the lead offers to switch that teammate to Claude Code or Cursor.

When you need Vowline

When several agents change the same product and you can no longer tell what still works. Vowline answers that question with checks, not with an agent saying “done”.

How to start

Also: Vowline and Cursor · Vowline and Claude Code

Codex is a product of OpenAI. Vowline is not affiliated with OpenAI and runs Codex with your own login.