The more I use Codex and Claude for real work, the more often I check their weekly limits. A remaining percentage tells me where I am, but it does not answer the question I actually have: how much room do I have for today?
The same percentage means something different just after a reset and just before the next one. I wanted a reference point for deciding whether to start a heavy implementation, move research to another day, or spend capacity on an exception that matters now.
That is why I built QuotaTempo, a macOS menu bar app that places remaining Codex and Claude capacity beside the target pace for the current point in the week. It is an independent open-source project, not an official OpenAI or Anthropic product.
As of September 11, 2026, the available build is QuotaTempo 0.1.0 Public Beta (RC19). This is not a promise that the app will reduce usage or improve productivity. It is a small tool for making one weekly allocation decision with more context.
Remaining capacity does not tell me what I can use today
I used to look at the remaining number and react to it in isolation. A high number felt comfortable. A lower one made me hesitate, even when the work in front of me was time-sensitive.
The missing piece was time. If a weekly limit has just reset, a high remaining value is expected. Near the next reset, the same value may mean there is still room to use. A value that looks generous can also hide the fact that I have consumed capacity faster than an even weekly pace.
This became more noticeable as Codex took on implementation and operational work while Claude helped with review and alternative perspectives. Before choosing a model or opening another task, I wanted to know whether the current week was ahead of or behind the pace I had intended.
QuotaTempo does not turn that decision into a rule. It gives me a comparison that the remaining number alone cannot provide.
QuotaTempo compares W with target pace P
The main display uses two values:
Wis the weekly capacity currently available from the recognized provider state.Pis the target amount that would remain now if the available capacity were used evenly until the next reset.- The arrow and
ptsshow the difference betweenWandP.
When W is above P, I have more room than the even-use plan suggests. When it is below P, my usage has moved ahead of that plan. Neither state is an instruction.
A release, an incident, or a deadline may justify using more today. A quiet day does not require me to consume spare capacity simply because it is available. I read the display as a pace chart, not as a speedometer or an automatic scheduler.
The screen below is a synthetic fixture from the public repository. It illustrates the interface and does not contain a real user’s quota values.
Full, Compact, and Icon Only
QuotaTempo offers three menu bar modes. Full keeps the provider values, target pace, and difference visible together. Compact reduces the amount of text when I only need the short comparison. Icon Only leaves the details inside the menu.
I made Icon Only the initial mode because menu bar items can disappear behind the notch on some Macs. On first launch, a separate onboarding window explains the three options so that a hidden menu bar label does not look like a failed launch.
That detail came from a practical failure rather than a feature list. If the app opens but its status item is hidden, the first experience is already confusing. The onboarding screen exists to prevent the interface from failing before the numbers matter.
This image is also a synthetic fixture, not a capture of a live account.
Install the public beta safely on Mac
The September 11 Public Beta supports Apple silicon Macs running macOS 14 or later. To display a provider, the corresponding Codex app or CLI, Claude Desktop, or Claude Code must already be signed in and recognized on the Mac.
The release ZIP is signed with a Developer ID certificate and notarized by Apple. I still verify the downloaded artifact before opening it:
- Download
QuotaTempo-0.1.0-rc.19-macOS.zipandSHA256SUMSfrom the RC19 release into the same folder. - In Terminal, move to that folder and run
shasum -a 256 -c SHA256SUMS. - After the checksum reports
OK, double-click the ZIP in Finder. - Drag
QuotaTempo.appto Applications in Finder. - Open QuotaTempo from Applications.
If macOS rejects the app, I do not recommend bypassing Gatekeeper. Stop and compare the file with the release information and the user guide.
During development, I once opened a downloaded copy before moving it correctly into Applications, and it ran under App Translocation. I am not treating Terminal use as the cause. For final acceptance, I downloaded the official release again on another Mac, expanded the ZIP in Finder, moved the app to Applications, and opened the copy located there.
Codex and Claude do not expose identical data
QuotaTempo presents Codex and Claude in a consistent layout, but the underlying observations are not identical. It reads only recognized local provider state needed for the display.
For Codex, the app uses the recognized path associated with the official Codex app or CLI. For Claude, it uses recognized usage information left by Claude Desktop or Claude Code. The available fields can differ.
Claude Desktop may expose weekly remaining capacity without a confirmed reset time. In that case QuotaTempo does not invent a target pace; it shows Reset time unavailable. Using Claude Code once may create the corresponding local observation, but that depends on the provider’s current behavior.
W? means the current weekly value is still being refreshed. P≈ marks a plan estimated under limited conditions. If a provider changes its interface or local data shape and QuotaTempo no longer recognizes it, the app is designed to show the reason and stop rather than make stale or guessed data look valid.
The degraded-state image is a synthetic fixture showing those boundaries.
Keep usage data local
A capacity display can become sensitive very quickly, so I defined what QuotaTempo would not collect before adding more features.
According to the project’s Privacy document, the app has no first-party telemetry, analytics, account system, advertising identifier, remote license check, or operator server. It does not read provider credentials, browser cookies, Keychain items, prompts, transcripts, or session bodies. Raw provider responses are not stored; only normalized state needed by the interface remains on the Mac.
Local does not mean every related network connection is zero. Provider processes that are separate from QuotaTempo may communicate according to OpenAI’s or Anthropic’s own behavior and terms. The boundary I can describe is what this app reads, stores, and sends—not everything another installed provider may do.
The same rule applies to diagnostics. A public issue should not include a personal quota percentage, reset timestamp, credential, prompt, transcript, private path, provider file, or raw response. The beta feedback form does not ask for those items.
Test whether it changes decisions for seven days
I do not need to watch QuotaTempo continuously. For the first test, I check it once a day at roughly the same time for seven days.
I record the decision, not the private quota number:
- move work forward;
- continue roughly as planned;
- hold a heavy task for later;
- make an exception for work that matters today.
Beside that decision, I add one short label such as implementation, research, review, or writing. After seven days, the useful question is not whether usage fell by a certain percentage. I have not measured that. The question is whether comparing W with P changed when I began, postponed, or prioritized substantial work.
QuotaTempo does not increase a limit, route prompts, recommend a provider, or work around quota rules. It leaves the final allocation decision with the person who knows the deadline and the consequence of waiting.
Conclusion
Having access to Codex and Claude did not automatically give me a stable weekly plan. Remaining capacity described the current state, but I still needed a reference point for deciding what to do today.
QuotaTempo adds that reference point by comparing weekly remaining capacity with the target for the current moment. It deliberately stops short of making the decision for me.
The Public Beta has no usage deadline, but it does not include automatic updates, and the formal release date, future features, delivery model, and pricing are undecided. Check GitHub Releases for the current build rather than assuming this article will stay current forever.
Try it for seven days and record whether your decisions change. If it stops at an unavailable state, report the point of failure through Public Beta feedback without attaching private usage or account data.
Related reading
- How I interpret Codex’s banked rate-limit reset
- How I audit a local Codex environment before changing it
- Using codex-healthkit for a metadata-only local environment check