How I keep Codex agent sessions under control without losing drafts
Developers juggling multiple Codex agent sessions lose track of which runs finished and lose drafts when they switch model providers. Codex's updated command center addresses this by marking unread results and preserving drafts across backend switches.
What it is
The Codex CLI agent command center, now usable without a daemon via codex agents --no-daemon, tracks session state across the terminal. It flags unread results and preserves drafts when switching connected backends.
- Launch with codex agents --no-daemon
- Unread results marked in command center and session picker
- Drafts survive switching between connected backends
Applied case
A developer running Codex against multiple providers previously lost their in-progress prompt when swapping backends mid-session. Now they open the session picker, see which finished runs they have not read yet, and switch to a Vertex AI backend configured beforehand through the gcp_setup_wizard while the draft stays intact. If a chat cannot resume after switching, they can still read the transcript and spin up an editable prompt to continue in a fresh chat. Two-minute CTA: run codex agents --no-daemon and check for unread badges on your finished runs.
- Unread result badges stop you missing completed agent runs
When to use it
Use when you run multiple agent sessions or juggle providers in one day. Skip when you work in a single uninterrupted session with one backend; this update does not add cross-session memory or recover chats that cannot resume, it only keeps transcripts readable.
Read more
Recommended for you
- WorkflowVercel
How I keep agent loops fast without waiting on slow models
Agent workflows stall when each step waits seconds for model output, so throughput becomes the real bottleneck.
- WorkflowClaude Code
Why does your CI agent still stall on permission prompts?
Unattended Claude Code turns hang whenever a tool call waits for a permission prompt nobody answers, breaking scripts and subagent workflows.
- WorkflowVercel
When to use a vision coding agent instead of a plain LLM?
Text-only models miss detail in screenshots and charts, and drift on long-running agent tasks across a large codebase.
- WorkflowVercel
How I get usage credit without juggling a dedicated Token Pass key
Token Pass credit now applies automatically to AI Gateway requests, no special key required, with fallback to team credits when it runs out.
Enjoyed this article?
Subscribe for new articles. No spam. Unsubscribe anytime.