agents · review · first seen Nov 6, 10:33 UTC

OpenAI adds Codex command center and backend switching

Codex 0.167.0 shipped an agent command center, backend switching in the terminal session picker, and local image previews on supported terminals. Builders should check whether these changes reduce friction in work that crosses sessions or connected backends.

L0 · primary1 source5 claims checked4 min read

What shipped

Codex 0.167.0 shipped an agent command center that opens with codex agents --no-daemon. The command center and session picker now mark unread results. That gives terminal users a place to look for work awaiting attention, but the release notes do not say the command center replaces an existing workflow. Open it and check whether the unread markers identify results you have not read. [1][2]

The terminal session picker shipped switching between connected backends while retaining drafts. Saved chats also stay visible after switching model providers. These are distinct behaviors: one preserves text you are composing; the other preserves access to saved conversations. The release notes add a limit to the latter. If a chat cannot resume, you can read its transcript and prepare an editable prompt to continue in a new chat. Visibility is not a guarantee of resumption. [1][2]

On supported terminals, Codex shipped previews of local image attachments while composing. The release notes also describe opening viewed local images from clickable transcript paths. If you attach images as prompt context, check whether you can inspect an attachment before sending the draft. The stated support condition matters: the source does not say that every terminal renders previews. [1]

Why it matters if you build

If you watch several agent sessions, an unread marker in both the command center and session picker offers a cue for where to look next. Compare those markers with the sessions whose results you expect to review. The evidence supports a navigation aid, not a claim about notifications, prioritization, or how quickly results arrive. There is no measured throughput gain here to put into a rollout case. [1][2]

If you alternate among already connected backends, retained drafts address a specific interruption: losing text you were composing when you switch. Saved-chat visibility addresses a separate concern: losing sight of a prior conversation after changing model providers. Test each behavior separately with a draft and a saved chat. Do not infer that a visible chat can resume under another provider; the release notes describe a transcript and editable-prompt route when it cannot. [1][2]

The image preview is worth testing if your prompts depend on choosing the right local attachment. Use the terminal you actually work in, because support is qualified. A preview lets you inspect the attachment while composing, but the supplied evidence gives no terminal compatibility list or performance measurements. Likewise, backend switching applies to connected backends; the claim does not say switching connects a new one for you. Check those boundaries before making either feature part of a team routine. [1][2]

How to try it today

1. Open Codex 0.167.0 with codex agents --no-daemon and inspect the command center.
2. Check the command center and session picker for unread-result markers.
3. Compose a draft, switch between connected backends in the terminal session picker, and check that the draft remains.
4. Switch model providers and check whether a saved chat stays visible. If it cannot resume, inspect its transcript and editable continuation prompt.
5. Attach a local image while composing on a supported terminal and check its preview.

What we checked

Every claim below was walked down the source ladder. The rung is how close the source is to the thing itself; the tense is the source's own.

1. shipped L0 · primary

   Codex 0.167.0 adds an agent command center opened with codex agents --no-daemon.

   per GitHub releases · openai/codex (L0)
2. shipped L0 · primary

   The command center and session picker mark unread results.

   per GitHub releases · openai/codex (L0)
3. shipped L0 · primary

   The terminal session picker lets users switch between connected backends while keeping drafts.

   per GitHub releases · openai/codex (L0)
4. shipped L0 · primary

   Codex previews local image attachments while composing on supported terminals.

   per GitHub releases · openai/codex (L0)
5. shipped L0 · primary

   Saved chats stay visible after users switch model providers.

   per GitHub releases · openai/codex (L0)

What is not confirmed

The supplied excerpt does not establish which terminals support image previews, whether any particular saved chat will resume after a provider change, or whether these changes reduce time or cost. It gives no measured reliability for unread markers or draft retention. Treat the listed behavior as shipped features to check in your setup, not as measured outcomes. The qualifications are supported terminals for image previews and connected backends for switching; neither supplies a broader compatibility guarantee. [1][2][3][4]

Who should act, who should wait

Act today if you already use Codex in a terminal to track agent sessions or switch among connected backends. Try the command center and picker, then verify draft retention and saved-chat visibility with your own sessions. If local images are part of your prompts, check previews in your actual terminal. Wait before depending on previews across unspecified terminals or assuming cross-provider chat resumption: neither is established here. Test the shipped interface changes, but keep unsupported compatibility and outcome claims out of the decision. [1][2][3][4][5]

Sources

• 0.167.0 — GitHub releases · openai/codex L0 · primary