Home / Guides / Codex Usage
How to Check Your Codex Usage and Rate Limits
You can check Codex usage three ways: run /status inside the Codex CLI for live 5-hour and weekly limits, open platform.openai.com/usage for historical tokens and cost, or use a menu-bar tracker for always-on monitoring without leaving your editor.
Last updated May 2026 · Updated for OpenAI's April 2026 token-based rate-card change · By Soren Starck, indie dev in Lyon and maker of SessionWatcher
Understanding Codex Usage Limits
If you use OpenAI Codex for coding, you're subject to two stacked rate limits: a rolling 5-hour window and a weekly window. Both are now measured in tokens consumed (post-April 9, 2026), so an agent that does a lot of reasoning burns through your budget faster than the message count would suggest.
The problem? There's no built-in warning system before you hit either cap. You're coding, in flow, making progress, and suddenly Codex stops responding. Now you're stuck waiting, with no clear indication of when capacity frees up.
Method 1: Run /status inside Codex CLI
The fastest free check is the built-in slash command. Inside an active Codex CLI session, type:
/statusYou'll get the remaining tokens for the rolling 5-hour window and the weekly window, plus your current model and plan tier.
Limitations: /status only works while Codex is open and active: it's a snapshot, not a watch. You also have to remember to run it. The numbers can be slightly stale on the first invocation of a session (an open issue tracked in the Codex GitHub repo).
Method 2: OpenAI Dashboard
The dashboard is the canonical source for historical billing data:
- Go to platform.openai.com/usage
- Log in to your account
- Filter by date range and model to see Codex-specific consumption
- Cross-reference with the Codex rate card for credit-to-token mapping
Limitations: The dashboard shows historical tokens and cost, but it doesn't tell you where you are in the rolling 5-hour window right now, or how close you are to the weekly cap. The data also lags by a few minutes, which makes it useless for catching yourself before a lockout.
Method 3: Free CLI trackers (ccusage, etc.), and why they don't solve Codex
You may have seen open-source usage trackers like ccusage, Claude-Code-Usage-Monitor, or Claude Usage Tracker. They're great tools, but only for Claude.
All three read Anthropic Claude Code's local ~/.claude/projects/*.jsonl session files. They do not read Codex session data — different session format, different API, different vendor. For Codex you need a tool that actually parses Codex sessions.
But for Codex specifically, your free options collapse back to /status and the OpenAI dashboard. To get an always-on Codex tracker without writing your own OpenAI Usage API script, you need a tool that actually parses Codex session data. Today that means either rolling your own or using SessionWatcher's Codex build.
Method 4: SessionWatcher (always-on, in your menu bar)
Here's the friction the free options share: /status requires you to context-switch into Codex and remember to ask. The dashboard requires a browser tab and a login. Neither warns you before a lockout. Both interrupt flow at exactly the moment you're paying to avoid interruption.
SessionWatcher for Codex is a native macOS menu bar app that solves this by being always visible. Glance up to see your live token count, cost, the 5-hour countdown, and the weekly window. No commands, no logins.
- Real-time tokens. Exact consumption in the current rolling window.
- 5-hour window countdown. See when capacity frees up.
- Weekly window tracking. The second cap most tools ignore.
- Cost in dollars. Credits-to-cost translated for you.
- macOS notifications. You're warned before you hit the cap, not after.
- 5 tools in one app. Claude Code, Codex, Copilot, Cursor, Gemini CLI.
SessionWatcherStop finding out at 2am.
Just glance up.
Native macOS menu bar app. Track Claude and Codex usage, costs, and rate limits in real-time.
“Fast, simple, and does exactly what it should. Definitely worth it.”
@nicojerome on GitHub
macOS 14+. $2.99 one-time purchase.
Side-by-side: Dashboard vs /status vs ccusage vs SessionWatcher
| Capability | OpenAI Dashboard | /status | ccusage (CLI) | SessionWatcher |
|---|---|---|---|---|
| Tracks Codex | Yes | Yes | No (Claude only) | Yes |
| Live 5-hour window | No | On demand | N/A | Always visible |
| Weekly window | No | On demand | N/A | Always visible |
| Pre-lockout warning | None | None | N/A | macOS notification |
| Multi-tool (Claude, Codex, Copilot, Cursor, Gemini) | Codex only | Codex only | Claude only | All five |
| Setup | Login each time | Built in | npm + CLI | 10-second install |
| Cost | Free | Free | Free | $2.99 once |
| Open Dashboard → | Built into Codex CLI | ccusage on GitHub → | Get for Mac |
Editorial note: ccusage is excellent for Claude-only users. It doesn't support Codex (different vendor, different session format), which is why this page exists.
How the rolling 5-hour window works (post-April 2026)
The 5-hour window is rolling, not fixed. Tokens you consumed at 9:00 AM begin freeing up at 2:00 PM. Here's a token-based example under the new system:
- At 9:00 AM, an agent run with heavy reasoning consumes 50% of your 5-hour token budget
- At 11:00 AM, a refactor pass eats another 40%, putting you at 90%
- At 11:30 AM, one more reasoning-heavy request and you're locked out
- At 2:00 PM the 9:00 AM tokens roll off and you regain ~50% capacity
- The weekly window keeps counting in the background regardless
Because tokens (and reasoning time) are now the unit, two sessions of identical “message count” can have wildly different impact on your limit. SessionWatcher does the math live; /status shows it on demand; the dashboard tells you after the fact.
Frequently Asked Questions
How do I check my Codex usage?
Three options: run /status inside Codex CLI for live numbers, open platform.openai.com/usage for historical data, or use SessionWatcher's menu bar Codex tracker for always-on monitoring.
What is the Codex 5-hour rate limit window?
A rolling 5-hour window plus a weekly window. Both are now metered by tokens consumed (post-April 9, 2026). The longer the agent reasons, the more of your 5-hour limit it consumes. Message count is no longer a reliable proxy.
Why did my Codex stop working mid-session?
You hit either the 5-hour rolling cap or the weekly cap. Codex stops responding without a clear warning. SessionWatcher prevents this by showing live usage and sending a macOS notification before you reach either cap.
Is there a free Codex usage monitor?
/status is built into Codex CLI and free. The OpenAI dashboard is free but lags. Free open-source CLIs like ccusage are Claude-only and don't track Codex. For an always-on Codex tracker, SessionWatcher is $2.99 one-time.
Does ccusage work with Codex?
No. ccusage parses Anthropic Claude Code's local JSONL files and is Claude-only. The same goes for Claude-Code-Usage-Monitor and Claude Usage Tracker. For Codex you need /status, the OpenAI dashboard, a custom OpenAI API script, or a tool that reads Codex session data directly.
Is there a Codex monitor app for macOS?
Yes. SessionWatcher for Codex is a native macOS menu bar app. $2.99 one-time, or $7.99 for all five tools. Zero configuration, works in 10 seconds.