The real Codex, on your plan
Every Codex agent in AI Bridge is the codex command you would run yourself, in a real terminal, using your own ChatGPT plan or API key. AI Bridge adds a place for it to live, not a layer between you and OpenAI.
It never edits your ~/.codex configuration: what it needs to know a session’s state is passed on that launch’s command line, so your setup in your own terminal stays exactly as it was.
Know when Codex is waiting on you
Codex stops to ask before it runs a command it is not sure of. In a terminal tab you notice when you happen to look. In AI Bridge the session says it is waiting on you, in the sidebar and in the Inbox, beside everything else that needs you. See how that works across several agents.
Side by side with Claude Code
Give Codex and Claude Code the same problem and watch both on the Grid, a wall of live sessions you arrange yourself. Or give each the work it does best: Codex in one workspace, Claude Code in another, Grok and OpenCode where they fit.
Organised by the job
A Codex session starts in a workspace, an area of responsibility with its own instructions, context files and repos, so it begins knowing the rules of the job rather than nothing. Put it to sleep at the end of the day and wake it tomorrow where it stopped.
Codex in a Slack channel
A Codex agent can be bound to a Slack channel: each message becomes a job you approve before it starts, and each reply waits for you to send it. It can also work beside your clusters and your databases.
