Can BalkayaClaude Code and Codex can both be busy while answering very different questions about usage. A token total tells you how much work the models processed. A plan window tells you how much allowance remains. Your subscription invoice tells you what you paid. Keep those three numbers separate before comparing agents or changing a plan.
This guide is published by Codz. It shows a Mac workflow for following both agents together; you can also inspect each provider’s own usage tools independently.
Decide which number you need
| Question | Number to use | What it cannot tell you |
|---|---|---|
| How much did the agents process today? | Recorded tokens, split by provider and model | Whether the work was useful or correct |
| Am I about to hit a limit? | The provider’s plan window, remaining allowance and reset | Your monthly bill |
| What did I pay? | Your subscription invoice or provider-reported API charge | The list-price value of subscription tokens |
| How would this usage be priced on an API? | An explicitly labelled API list-price estimate | What your subscription actually cost |
An API-value estimate can be much larger than a subscription invoice. That is a pricing comparison, not money you have been charged or savings Codz has delivered. Missing model prices should remain missing rather than silently become zero.
Put Claude Code and Codex on the same timeline
- Install Codz on a Mac running macOS 14 or later.
- Use the Claude Code and Codex accounts you already use. Codz reads their local session records; an account with Codz is not required for the free local workflow.
- Open Usage from the account menu and choose the same period for both agents. Give the initial history import time to finish before drawing conclusions.
- Compare providers and models over that period. Look for a changed workload or a different model before assuming that one agent became more expensive.
The live example uses illustrative data. Your own totals come from the records on your Mac.
Codz reads Claude Code’s session transcripts and Codex’s session logs. Those records support token totals and an estimate of API value; they do not report your subscription invoice. The usage reference explains the source and meaning of each figure, including account-reported charges from other providers.
Watch the allowance while you work
A daily total is useful for looking back. During a task, the next reset and the window running out first are often more useful. Open the menu-bar panel to check the plan windows alongside the agents currently working.
A provider that has not reported a limit cannot be treated as having unlimited allowance.
When a window is nearly spent, finish or pause a well-defined task before moving work to another provider. Carry over the goal, decisions, files changed, tests run, and the next step. A new agent still needs to review that context; changing providers does not guarantee the same answer.
See plan limits for the supported windows and handoffs for continuing a task with another agent.
Compare work, then compare cost
For a useful weekly review, pick one project and ask:
- Did the workload change, or did the same kind of work consume more tokens?
- Which models performed the work? Did a large context or repeated retries explain the increase?
- Were recorded charges and estimated API value kept separate?
- Did the changes pass review and testing?
Token totals alone are a poor productivity score. A short failed change is not better than a longer correct one, and a cached input token is not interchangeable with a new output token for pricing.
Know the free and paid boundaries
The free Mac app includes the last seven days in Usage, live plan limits, Tokens Today and codz usage. Pro adds longer history and deeper reporting; synced web and team analytics also require Pro. Check the current plans before choosing a reporting workflow.
Codz does not replace your Claude or OpenAI subscription, and its price does not include those providers’ model usage. If you only need one provider’s remaining allowance, that provider’s own usage view may be enough. Codz is useful when you want the work of several agents visible together on your Mac.
If the totals look wrong
First check the date range and whether history is still being read. If macOS has blocked access to the providers’ records, follow Codz’s permission notice; do not grant access to unrelated folders just to make a chart populate. If tokens exist but a cost is unavailable, inspect the explanation next to the number.
The troubleshooting reference and usage analytics documentation cover missing history, permissions, and the difference between reported charges and estimates.
See your agents’ usage together
Start with the free Mac app. Longer history and synced analytics are available with Pro.
macOS 14 or later · Apple silicon and Intel. Installation guide
Sources and review
Written by Can Balkaya. Published by Codz. Product details were checked on November 3, 2026; follow the sources for current requirements and pricing.