📊 pi-codex-usage — Codex Usage Status for Pi

[!WARNING] @narumitw/pi-codex-usage is deprecated and will be replaced by @narumitw/pi-usage. Do not load both extensions at the same time. To migrate one Pi installation, run:

pi uninstall npm:@narumitw/pi-codex-usage
pi install npm:@narumitw/pi-usage

See the pi-usage migration guide for behavior changes.

@narumitw/pi-codex-usage is a native Pi coding agent extension that adds /codex-status, a command for showing ChatGPT Codex subscription usage from inside Pi.

Use it when you want a quick Codex-style usage summary without leaving Pi or requiring Codex CLI to be installed.

✨ Features

• Adds a /codex-status command to Pi.
• Shows Codex usage windows, reset times, credits, and earned usage-limit resets.
• Labels each window from its reported duration (for example, 5-hour or weekly).
• Displays additional usage buckets when the Codex backend returns them.
• Reads the authoritative available reset count from the current Codex usage contract.
• Automatically shows a compact statusline item while the current Pi model uses openai-codex.
• Uses Pi's own OpenAI Codex subscription auth first.
• Falls back to codex app-server --listen stdio:// only when Pi auth is unavailable.
• Caches results briefly to avoid repeatedly calling the backend.
• Supports --refresh to bypass the in-memory cache.
• Works as an independently installable npm Pi extension package.

📦 Install

pi install npm:@narumitw/pi-codex-usage

Try without installing permanently:

pi -e npm:@narumitw/pi-codex-usage

Try this package locally from the repository root:

pi -e ./extensions/pi-codex-usage

🚀 Usage

/codex-status
/codex-status --refresh
/codex-status --no-statusline
/codex-status --clear-statusline
/codex-status --timeout 30

Example output:

  >_ OpenAI Codex Usage

Visit https://chatgpt.com/codex/settings/usage for up-to-date
information on rate limits and credits

  5h limit:                    [█████████████░░░░░░░] 64% left (resets 13:57)
  Weekly limit:                [████████████░░░░░░░░] 62% left (resets 14:37)
  GPT-5.3-Codex-Spark limit:
  5h limit:                    [████████████████████] 100% left (resets 19:16)
  Weekly limit:                [████████████████████] 100% left (resets 00:10 on 21 May)

  Usage limit resets:          2 available

📊 Statusline behavior

When the selected Pi model provider is openai-codex, pi-codex-usage refreshes a compact statusline item automatically:

codex 59% 5h 61% wk
codex spark 100% 5h 100% wk

@narumitw/pi-statusline adds the default 📊 icon unless configured otherwise. The statusline value uses the cached usage snapshot and refreshes every five minutes while the current model remains openai-codex. Window labels come from the duration reported by Codex, with 5-hour/weekly fallbacks for older responses that omit it. When the selected model has its own returned usage bucket, such as gpt-5.3-codex-spark, the statusline switches to that bucket instead of the default codex bucket. Switching away from an OpenAI Codex model clears the item.

Use /codex-status --no-statusline for a one-off notification without updating the statusline, or /codex-status --clear-statusline to clear the item manually.

🔐 Auth behavior

pi-codex-usage tries usage sources in this order:

1. Pi's openai-codex provider auth through the Pi extension API.
2. Codex CLI app-server fallback when Pi auth cannot provide usable subscription auth.

This means Codex CLI is optional. Users who already use a Pi OpenAI Codex model or have logged in to Pi with ChatGPT Plus/Pro subscription auth can use the direct Pi-auth path.

The direct /wham/usage response can include the snake_case rate_limit_reset_credits summary. Current Codex app-server responses expose the same authoritative count as camelCase rateLimitResetCredits and may also include capped detail rows. The extension accepts both forms and keeps the compact statusline focused on rate-limit windows.

The extension does not read Pi or Codex auth files directly, and it does not expose bearer tokens in error messages.

🚧 Limitations

• OpenAI API keys are not ChatGPT Codex subscription auth and do not expose this quota.
• Usage data is a snapshot. Statusline and command results are cached for five minutes unless --refresh is used.
• The fallback path requires Codex CLI to be installed and logged in.
• /codex-status reports earned usage-limit resets but does not redeem them.

🗂️ Package layout

extensions/pi-codex-usage/
├── src/
│   ├── codex-usage.ts  # Pi entrypoint and cache/lifecycle orchestration
│   └── *.ts            # Package-local query, RPC, normalization, and format modules
├── README.md
├── LICENSE
├── tsconfig.json
└── package.json

Only codex-usage.ts is a Pi entrypoint; the other source modules are internal. The package exposes its Pi extension through package.json:

{
  "pi": {
    "extensions": ["./src/codex-usage.ts"]
  }
}

🔎 Keywords

Pi extension, Pi coding agent, Codex usage, Codex status, ChatGPT subscription usage, rate limits, TypeScript Pi package, npm Pi extension.

📄 License

MIT. See LICENSE.