Codex CLI · 3 min read

Reusable prompts in ten numbered /presets slots

Codex CLI 0.164.0 adds ten numbered /presets slots for reusable prompts and slash commands. It also adds inline arguments for /model and /permissions, image previews in the fullscreen transcript, and a default-on Amazon Bedrock setup wizard.

Key takeaways 5
  • Codex CLI 0.164.0 adds ten numbered /presets slots where you can save reusable prompts and slash commands, with a keyboard menu for quick access.
  • /model and /permissions now accept inline arguments, so you can set model, reasoning effort, speed, and permission profiles in a single command.
  • The fullscreen transcript can preview viewed and generated images on terminals that support Kitty or Sixel graphics.
  • The Amazon Bedrock setup wizard is now on by default, and API key entry is grouped with the other AWS sign-in methods when credentials are detected.
  • Enterprise MCP connections can authenticate with eligible Codex account grants or a dedicated sign-in, and credentials are masked during supported remote Bash commands and Zsh snapshot sessions.

Most long Codex CLI sessions repeat themselves: the same review prompt, the same slash command, the same model and permission setup typed out again. This week’s release goes after that repetition with saved prompt slots and shorter commands for switching settings. It also adds image previews in the fullscreen transcript and a few changes for people who run Codex against AWS or enterprise MCP servers.

This update covers Week 43 (October 23, 2026) and version 0.164.0: one release with 14 changelog entries. Four main features and two smaller changes made it into this summary.


Ten slots for the prompts you keep retyping

Ten numbered slots. Codex CLI now has /presets, a set of ten numbered slots where you save reusable prompts and slash commands. Each slot holds one of them, so your most common instructions sit in a fixed place instead of in your shell history or a notes file.

A keyboard menu for access. A keyboard menu gives quick access to the slots. The idea is that recalling a saved prompt costs a keystroke or two instead of a paste.

Slash commands belong in the slots too. You can store slash commands there as well as prompts. If you have a command sequence you run at the start of every session, a numbered slot is a natural home for it.

What to check. The release notes give no details on how slots are edited, whether they persist across projects, or which shortcut opens the menu. Open /presets in 0.164.0 and see how it behaves before you plan a team convention around it. (Ten is a small number, so decide which prompts deserve a slot.)


Changes for everyday work

Inline arguments for /model and /permissions. You can now pass arguments directly to /model and /permissions to set the model, reasoning effort, speed, and permission profile. The release notes do not list the argument syntax, so check it in the CLI before you script anything around it. The practical effect is that a settings change no longer has to go through a picker.

Image previews in the fullscreen transcript. Images you viewed and images Codex generated can now be previewed in the fullscreen transcript. This needs a terminal that supports Kitty or Sixel graphics. On a terminal without either, expect no preview.

Amazon Bedrock wizard on by default. The Amazon Bedrock setup wizard is now enabled by default. When credentials are detected, API key entry is grouped alongside the other AWS sign-in methods, so you pick your method in one place.


Technical notes

Enterprise MCP authentication. Enterprise MCP connections can authenticate using eligible Codex account grants or a dedicated sign-in. That includes OAuth clients that require secrets. Which grants count as eligible is something your admin or the documentation will have to answer.

Credential masking. Credentials are now masked during supported remote Bash commands and Zsh snapshot sessions. This includes credentials from custom providers. The word ‘supported’ matters here: the notes do not say which setups fall outside it, so do not treat masking as a blanket guarantee.


What to take away

If you work mostly in the terminal:

  • Open /presets and move your most repeated prompts and slash commands into the ten slots.
  • Try inline arguments with /model and /permissions to switch model, reasoning effort, speed, or permission profile in one step.
  • If your terminal supports Kitty or Sixel graphics, open the fullscreen transcript and check the image previews.

If you run parallel sessions:

  • This release has no feature aimed squarely at parallel sessions. Inline /model and /permissions arguments may still make it quicker to give each session its own model and permission profile, so test that.
  • Check whether the numbered /presets slots are shared across your sessions or tied to one. The release notes do not say.

If you administer integrations or company deployments:

  • Review how your enterprise MCP servers authenticate. Eligible Codex account grants or a dedicated sign-in are now options, including for OAuth clients that need secrets.
  • Check which of your remote Bash and Zsh snapshot setups count as supported for credential masking, especially if you use custom providers.
  • The Amazon Bedrock wizard is now on by default, so expect users with detected AWS credentials to see the grouped sign-in options.

Sources: the Codex CLI changelog, version 0.164.0 (October 23, 2026). The changelog entries cited here did not come with documentation links, so details such as /presets behavior and argument syntax are best confirmed in the CLI itself.

Patrick Zandl trains teams in Claude Code and agentic software development in Prague and writes about what actually works in real-world AI-assisted projects. Follow on LinkedIn

Read next

Get the next one by email

Every Tuesday: what changed in Claude Code, Codex and other coding agents, plus new articles. No spam.