Codex Analytics: What You Can Actually Measure
What you can actually track in Codex depends entirely on your plan and how you connect. If you're on Plus or Pro, usage visibility is pretty limited: you can see your current allowance status and reset time, but OpenAI's documented Plus and Pro surfaces do not provide a turn-by-turn subscription token log or the same historical token analytics available to eligible Enterprise workspaces. Eligible Enterprise members get much more detail in the ChatGPT Desktop app, including 7-day and 30-day token charts. Workspace admins on eligible Enterprise and Edu plans can see up to 120 days of organizational usage with date filters and CSV exports. And if you authenticate with your own OpenAI API key instead of a ChatGPT subscription, that is tracked through an entirely separate platform analytics dashboard.
What Codex Usage Data Is Measurable Across Plans?
What you can measure depends on your plan tier, your seat type, and your workspace role. Plus and Pro users get basic allowance status, reset times, and credit balances where applicable. Detailed token breakdowns and historical charts are limited to eligible Enterprise workspaces. Deep organizational metrics are in the Admin Console. And exact per-request token logs exist only when you pay for the Platform API directly.
| Plan / Surface | What You Can Measure | Detail Level | Where to Check | Availability | Notes & Limitations |
|---|---|---|---|---|---|
| Personal & Business | Active Quota Allowance | Remaining allowance / usage status | Codex CLI (/status), Web Settings | Plus, Pro, Business Standard | Shows active allowance draw; does not provide an itemized token invoice. |
| Personal & Business | Reset Information | Reset time and date when available | Codex CLI (/status), Web Settings | Plus, Pro, Business Standard & Premium | Indicates when rolling windows recover or weekly cycles reset. |
| Personal & Business | Working Session Memory | Percentage of context window free | Codex CLI (status bar / footer) | All CLI users | Measures session buffer before auto-compaction; distinct from account quota. |
| Personal & Business | Local Session Tracking | Quota delta since tool opened | Local Codex Usage Companion (codex-usage --web) | All users (Third-party tool) | In-memory only; auto-refreshes every 60s; no persistent external logging. |
| Enterprise Desktop | Historical Usage & Credit History | 7-day (7D) and 30-day (30D) trend charts | ChatGPT Desktop (Settings > Usage & billing) | Eligible Enterprise workspaces (credit-based billing, Desktop v26.812.10818+) | Monthly allowance, credits used, remaining allowance, next reset date where shown. |
| Enterprise Desktop | Itemized Token Breakdown | Uncached input, cached input, output tokens | ChatGPT Desktop (Settings > Usage & billing) | Eligible Enterprise workspaces (credit-based billing) | Categorizes historical token volume; not documented for Plus or Pro plans. |
| Enterprise Desktop | Usage Grouping & Filters | Filter by Surface, Model, Reasoning, Speed | ChatGPT Desktop (Settings > Usage & billing) | Eligible Enterprise workspaces (credit-based billing) | Breaks down credit consumption across execution surfaces and model configurations. |
| Enterprise Desktop | Top Consuming Chats | Ranked list of chats active in past 30 days | ChatGPT Desktop (Top Consuming Chats) | Eligible Enterprise workspaces (credit-based billing) | Displays per-chat lifetime credit usage; estimates main conversation stream only. |
| Enterprise Desktop | Estimated Cost Visibility | Dollar-denominated spend estimates | ChatGPT Desktop (where admin enabled) | Eligible Enterprise workspaces (credit-based billing) | Requires admin activation under Admin Console; unavailable on fixed Plus/Pro fees. |
| Enterprise Desktop | Personal Analytics Plugin | Natural language queries (@personal-analytics) | ChatGPT Desktop (Desktop plugin) | Eligible Enterprise & Edu workspaces (admin-enabled) | Read-only; self-metrics and aggregated workspace trends (trailing 30-day window). |
| Workspace Admin Console | Workspace Usage Totals | Active users, credits, tokens, message runs | Global Admin Console (Analytics) | Eligible workspace users with Codex admin role | Up to 120 days of Codex usage history with 1 to 6 hour refresh and date filtering. |
| Workspace Admin Console | Team & Code Activity | Lines of code, leaderboards, code reviews | Global Admin Console (Analytics) | Eligible workspace users with Codex admin role | User and agent leaderboards, lines of code, plugin calls, and skills used. |
| Workspace Admin Console | Data Export & Reporting | CSV export for selected date range | Global Admin Console (Overview) & Billing | Eligible workspace users with Codex admin role | CSV downloads for selected ranges; billing reports cover longer credit windows. |
| Platform API | Per-Request & Account Usage | Exact per-request tokens, custom date ranges | Platform Usage Dashboard & Usage API | Developers using custom OPENAI_API_KEY | Billed per token under API pricing; includes Usage API and CSV exports. |
| All Subscriptions | Per-Prompt Subscription Ledger | Turn-level token log for subscription turns | Not documented by OpenAI | None (for ChatGPT / Codex subscriptions) | OpenAI’s documented Plus and Pro usage surfaces do not provide this. |
What Usage Analytics Can Personal Plus and Pro Users See?
For Plus and Pro, this is the frustrating part: you do not get an itemized token history. What you can see is your current allowance status, reset time, and credit balance or recent usage where applicable. If you are running low or hit a limit, OpenAI points you to Settings or your account usage dashboard to check when your allowance recovers.
In an active Codex CLI session, running /status displays your remaining allowance and when your window resets. But neither /status nor your account settings provide a documented turn-by-turn token log or 30-day usage chart.
Subscription Quota vs. Working Session Context Memory
A common source of confusion in the CLI is seeing two different percentage numbers:
- Account Subscription Quota: Checked via
/statusor account settings, this tracks your consumption against plan limits across 5-hour rolling windows and weekly reset periods. - Session Context Memory: Displayed in the Codex CLI status bar or footer, this measures how much of the active model's context window remains free for your current terminal session.
That percentage is your context window, not your account quota. Reaching a context threshold or triggering auto-compaction just means your active conversation buffer is full—it does not mean your account-level Codex quota has been exhausted. For details on context limits, compaction triggers, and buffer sizing, see our Codex context window guide.
ChatGPT Business: Standard vs. Premium Seats
If you are on a ChatGPT Business workspace, what you see depends on your seat type:
- Business Standard Seats: Operate under documented plan limits, including the standard 5-hour rolling usage limit alongside weekly reset cycles.
- Business Premium Seats: Include 5x more Work and Codex usage than Standard and eliminate the 5-hour usage limit entirely.
Local Session Monitoring
Because documented Plus and Pro surfaces do not provide historical charts or token logs, keeping tabs on your burn rate during coding sessions usually means running /status over and over. If you want continuous visibility without checking manually, you can install the Codex Usage companion and run codex-usage --web.
The local dashboard refreshes your usage status every 60 seconds. Its session baseline and pace tracking live entirely in process memory, so that session history ends when you stop the process. The companion also does not send credentials or usage data to codexusage.dev.
What Analytics Do Eligible Enterprise Members Get in ChatGPT Desktop?
Enterprise gets much better history. For eligible Enterprise workspaces using credit-based billing, the ChatGPT Desktop app (version 26.812.10818 or later) includes a dedicated member-facing Settings > Usage & billing screen with actual 7-day and 30-day usage trends across Work and Codex.
Token Breakdown and Filtering
Eligible Enterprise members can review 7-day (7D) and 30-day (30D) trend charts covering monthly Work and Codex allowances, credits used, remaining allowance, and next reset dates where shown. You can filter and group consumption across four attributes:
- Surface: Compare credit usage across Desktop, CLI, Web, and integrated IDE extensions.
- Model: Track consumption across model families, including GPT-5.6 checkpoints (Sol, Terra, Luna) and GPT-6 Astra.
- Reasoning Effort: See how low, medium, and high reasoning configurations burn credits.
- Speed: Compare standard execution against accelerated modes where supported.
Beneath the trend charts, token volume is categorized into uncached input tokens, cached input tokens, and output tokens, alongside plugin and skill activity. Where enabled by workspace administrators under Admin Console > Usage limits > Workspace > Usage visibility and requests, members can also view estimated costs alongside credit figures. For screen layouts and navigation details, see our Codex usage dashboard guide.
Top Consuming Chats and What's Excluded
Under Usage & billing, the Top Consuming Chats section ranks locally available Work and Codex chats active within the past 30 days by lifetime credit use. Each entry shows its model, reasoning effort, speed, credits used, and estimated cost where available.
Keep in mind that current Work and Codex chat usage and monthly totals generally refresh about once per minute, while historical charts update less frequently. In addition, a chat's displayed total is an estimate for the main conversation stream only and may omit or incompletely reflect:
- Sub-agents and background tasks running outside the primary conversation thread.
- Image generation and separately billed tools carrying distinct billing rates.
- Background and system-generated activity.
- Delayed or incomplete reporting from client endpoints.
Personal Analytics Desktop Plugin
In eligible Enterprise and Edu workspaces where an administrator has turned it on, members can install the Personal Analytics plugin in ChatGPT Desktop.
It connects read-only to OpenAI's ChatGPT backends. You mention @personal-analytics directly in chat to ask natural language questions about your own activity and workspace averages. Documented query areas include turns, tokens, credits, models, task types, accepted code, code reviews, skills, plugins, and current usage limits where available.
Workspace summaries operate strictly over a fixed trailing 30-day window with minimum group-size thresholds to protect privacy. Personal Analytics only shows your own individual activity—you cannot inspect another member's stats.
What Analytics Can Workspace Administrators Access in the Admin Console?
If your role includes Codex analytics in an eligible Enterprise or Edu workspace, the Global Admin Console gives you up to 120 days of workspace usage history. You do not have to be a full workspace admin or global admin to see it: an owner can assign a custom role with “Allow members to administer Codex” enabled.
Admin Metrics and Refresh Timing
If your role includes Codex analytics, the Admin Console can show:
- User and Agent Activity: Active user counts, message runs, and user and agent leaderboards.
- Resource Consumption: Credits consumed, total tokens, usage breakdowns where supported, and estimated costs where available.
- Code and Tool Activity: Lines of code, plugin calls, skills used, and code-review metrics.
OpenAI documents up to 120 days of Codex usage history in the Admin Console, with data refreshing roughly every 1 to 6 hours. That makes it useful for tracking multi-month adoption, model utilization, and seat capacity. Keep in mind this Admin Console experience is documented for eligible Enterprise and Edu workspaces; Business workspaces do not have the same analytics interface.
Date Filtering and CSV Data Export
Unlike the fixed 7-day and 30-day views in the Desktop app, Admin Console Analytics supports custom date ranges rather than fixed intervals. From the Overview tab, admins and authorized roles can export usage data directly to CSV for any selected date range.
If you need credit history beyond that 120-day analytics window, OpenAI notes that older billing and credit records can be downloaded from the credit usage report in billing settings where applicable.
How Do Subscription and Workspace Analytics Differ from Platform API Analytics?
API usage is tracked separately from ChatGPT subscriptions. If you log in to Codex with your ChatGPT account, your usage draws from your subscription allowance or workspace credits. If you authenticate with your own OpenAI API key (OPENAI_API_KEY), your usage is billed per token under OpenAI Platform pricing and tracked in the Platform dashboard.
What you can measure in each environment is completely different:
- ChatGPT and Workspace Plans: You track consumption against plan allowances—like the 5-hour rolling windows and weekly resets on Plus and Pro, or credit pools on Enterprise. See our Codex usage limits by plan guide for tier allowances and when Codex usage resets for recovery schedules. OpenAI's documented subscription surfaces do not provide turn-by-turn token logs.
- OpenAI Platform API: Every API response can return exact token usage: prompt tokens, completion tokens, total tokens, and where supported cached-input or reasoning-token details. In the Platform Usage Dashboard, you can filter by project, user, and date range, export CSVs, or query the Usage API programmatically.
If you need exact per-request token accounting or programmatic API usage data, that requires API-key usage and Platform telemetry. ChatGPT subscription usage can still be monitored through tools such as the local Codex Usage companion, but it does not expose the same per-request token data.
Frequently Asked Questions About Codex Analytics
Can I see how many tokens a specific Codex prompt used on ChatGPT Plus or Pro?
OpenAI’s documented Plus and Pro usage surfaces do not provide an itemized per-prompt token count or turn-by-turn subscription token log. You can check your remaining allowance, reset timing, and credit balance (where applicable) using /status in the CLI or in your account settings, but individual turn-by-turn token counts are not exposed on those surfaces.
Does ChatGPT Business have a 5-hour usage limit?
It depends on your seat type. Business Standard seats operate under a rolling 5-hour limit alongside weekly reset cycles. Business Premium seats include 5x more Work and Codex usage than Standard and drop the 5-hour limit entirely.
What is the maximum history length for Codex analytics?
It depends on your role. Eligible Enterprise members can view 7-day and 30-day charts in the ChatGPT Desktop app. Workspace admins and users with Codex admin roles in eligible Enterprise and Edu workspaces can view up to 120 days of history in the Global Admin Console. Beyond 120 days, older credit data can be exported through billing credit reports where applicable.
Can workspace administrators see my prompt text in Codex Analytics?
Not in Codex Analytics. OpenAI's documented Admin Console Analytics views focus on aggregate metrics: active users, credits, tokens, message runs, leaderboards, lines of code, plugins, skills, and code reviews. The analytics documentation does not include prompt-text inspection as an analytics metric.
Does the local Codex Usage companion save my usage history?
No. The local dashboard refreshes your usage status every 60 seconds, but its session baseline and pace tracking live entirely in process memory. When you stop the process, that session data is gone. The companion also does not send credentials or usage data to codexusage.dev.