How to View Token Usage in Codex CLI: /usage, Fullscreen Mode, and Default Voice

AI Dev Tools — General
, , , , , ,
1

To see how many tokens you consume in Codex CLI, update to version 0.156.0 or later and type /usage: it opens an analytics panel (Codex usage) with your account usage, token totals, and plugin and skill activity. OpenAI launched it on September 22, 2026, in the same version that adds an optional fullscreen interface (/tui) and enables voice by default.

Release note: version 0.156.1 came out on September 23 and adds GPT-6 Sol and GPT-6 Luna to the model selector. Everything this note explains applies to both versions.

How do you view token consumption in Codex CLI?

Type /usage in any interactive session. According to the release notes and merged PRs, the panel includes:

  • A Summary tab with your account summary.
  • Token totals and history, with stacked charts.
  • Top chats: the conversations that consumed the most.
  • Plugin and skill activity, useful for detecting if an MCP server or a skill you installed and forgot about is silently consuming context.
  • Plan usage history, which the PRs describe as gated: it doesn’t appear in all accounts.

You can navigate it with your keyboard or mouse, and it comes with its own shortcuts help.

What you see depends on how you log in. The analytics PRs mention account-linked authentication and sections that depend on your plan. At the time of publishing this note, OpenAI’s documentation didn’t explain what /usage shows on each plan or if it works with API key authentication. If you’re missing a section, that’s likely why.

/usage or /status: which one do you need?

They answer different questions:

  • /status looks at your current session: the chat ID, how much of the context window you’ve used, and your rate limits. Use it while working, when you want to know if you’re about to hit compaction or a limit.
  • /usage looks at your account over time. Use it when the question is “where did my tokens go this week?”.

The same version also adds token counting and usage estimates per task in the agent command center, and quota warnings are now visible in the TUI instead of scrolling out of sight.

If what you’re looking for is to spend less tokens and not just measure them, at yoDEV we cover Headroom, which compresses context before it reaches the model.

How do you enable the fullscreen interface for Codex CLI?

Type /tui and choose fullscreen mode. The change applies from the next startup: you need to restart Codex to see it.

The fullscreen interface adds:

  • Search in the session transcript.
  • Mouse selection and right-click copy, both in the transcript and in the composer.
  • Clickable links in the responses.
  • Detail controls per activity, to expand or collapse each tool call instead of scrolling through it.

The classic interface remains the default. At the time of publishing this note, OpenAI hadn’t documented the config.toml key behind this option; /tui is the supported way to change it.

How do you disable voice in Codex CLI?

Press F8. It’s the shortcut that toggles voice conversations, which now come on by default in the terminal. The shortcut can be reassigned, but the release doesn’t indicate the name of the option.

Two related commands:

  • /voice settings lets you choose the voice for the next conversations.
  • Voice animations respect your system’s reduced motion preference.

The release notes don’t describe an option to disable voice permanently; F8 is the documented control.

Does voice work on Windows and Linux?

Yes. Audio runtimes are now included in the Linux and Windows builds, so you don’t have to install them separately. On Windows, Codex may show a notice about the Visual C++ runtime, which voice packages depend on.

What else changed in Codex CLI 0.156.0?

  • Worktrees enabled by default. You can create sessions in a worktree and filter tasks by status from the agent command center. If you already manage many sessions, combine it with our guide to keep your Codex CLI workspace clean with /archive.
  • Mermaid diagrams and equations render directly in responses.
  • Six new themes. Open the selector with /theme; your choice is saved in tui.theme inside $CODEX_HOME/config.toml.
  • Background server control. /daemon updates the local server and --no-daemon skips it.
  • Sandbox fixes. Isolation gaps close with inbound connections on Windows, privileged sockets on Linux/macOS, and writes through read-only handles on macOS.
  • Stability. The clipboard works in tmux and SSH sessions, streamed responses survive interrupted turns, and Plan mode restores when you resume a session.

How do you install or update Codex CLI?

On macOS and Linux, the standalone installer works for both installing and updating:

curl -fsSL https://chatgpt.com/codex/install.sh | sh

OpenAI’s documentation also offers installation with npm, Homebrew, and for Windows. Use the same method you used to install it.

What is Codex CLI?

Codex CLI is OpenAI’s code agent for the terminal. It reads your repository, edits files, and runs your local tools from a single session. You choose the model, reasoning effort, and permissions, and you can run it non-interactively in scripts and CI with codex exec. If you’re deciding between agents, check our comparison Claude Code vs Codex.