OpenAI published Codex CLI 0.167.0 on GitHub as tag rust-v0.167.0 at 2026-11-06T01:40:00Z (2026-11-06 09:40 Asia/Shanghai). The full changelog compares against rust-v0.166.0. Related reading: Codex CLI 0.166.0 Adds Transcript Selection and Queue Resume. The usual publish path is npm i -g @openai/codex@latest (npm latest listed 0.167.0).

Day-to-day CLI changes

The GitHub Release highlights codex agents --no-daemon, which opens an agent command center; the command center and session picker now mark unread results. The terminal session picker can switch across connected backends and keep drafts when switching. Supported terminals can preview local image attachments while composing, and viewed local images open from clickable paths in the transcript. Under “Sign in another way,” gcp_setup_wizard helps set up Google Cloud Vertex AI with Application Default Credentials or a credential file—a setup wizard, not a first-time Vertex provider (Vertex／gcp landed earlier in Codex CLI 0.165.0 Adds Vertex AI Provider and Memory v2 Defaults). /experimental can opt into macOS／Windows system proxy settings. While an approval prompt is open, patch details can expand directly in the transcript.

For Codex caps and ChatGPT-plan context on the site, see How Codex Usage Limits Work on Each ChatGPT Plan (October 2026). Coding-agent contrast: Claude Code vs Codex: What $20 Gets You and Which to Pick.

We haven't upgraded and exercised codex agents --no-daemon, cross-backend switching, local image previews, gcp_setup_wizard, or the experimental system-proxy path here.

Bug fixes (brief)

Release notes also call out saved chats staying visible after a model-provider switch; readable transcripts plus an editable prompt when a chat cannot resume; Ctrl+L clearing the display while earlier messages stay reachable via Ctrl+T; deleting text that exposes a leading ! still treating the line as an ordinary draft; network approvals closing when requests end and same-endpoint prompts serializing; realtime V3 long conversations saving completed speech; and automatic compaction accounting for retained output from failed／missing-usage responses.

Documentation／Chores in the same tag (app-server README network／prompt-user notes, openai-docs skill deprecation hints, DotSlash manifests, Linux musl allocator arenas) are docs／build work, not the product headline.

Product details can change; the latest GitHub Release wording applies. For reference only.