ChatGPT allowance explained: how connected app limits work

You connect an AI app to your ChatGPT account, try a few tasks and then notice your available usage dropping. Where did it go, and what does setting an app to 20% actually do?

Understanding the allowance helps you decide which tools deserve access and how much room to give them. Here is how the controls work, followed by examples you can use to plan your own setup.

What does “ChatGPT allowance” mean here?

Eligible Plus and Pro subscribers can let supported apps use their plan for AI requests. This draws on included Codex and ChatGPT Work usage. Signing in and granting plan usage are separate permissions. The app may charge its own fees. See OpenAI’s Sign in with ChatGPT guide.

Usage is affected by the model, task complexity, context and tools. A percentage cannot reliably translate into a fixed number of prompts. Check your dashboard for current limits and resets. OpenAI’s pricing guidance explains the variables.

For planning, ask which work the app will help you finish. A tool that repeatedly produces something you cannot use is a poor use of your allowance, even if each request seems small.

A 20% app limit does not reserve 20%

An app’s weekly cap is a percentage of overall weekly plan usage. It adds no allowance and reserves none. OpenAI describes the cap in its usage settings guide.

Consider this simplified example: represent a week’s allowance as 100 planning units. These are teaching units, not OpenAI credits, tokens or message counts. Set App A’s cap to 20%. Its ceiling is therefore 20 units.

If App A has used 5 units, it has used 5 of its possible 20. You have not automatically spent the other 15. If other work has already used 85 units, total usage is 90, leaving 10 overall. App A’s unused cap does not make another 15 units available.

The practical lesson is to check both the app ceiling and the remaining overall allowance before starting a substantial task.

Why an app can stop before its weekly cap

Plus users also share a five-hour usage limit across apps using their plan. An app can therefore have weekly room left while the shared shorter window is exhausted. That five-hour limit does not apply to Pro users. See OpenAI’s account and usage documentation.

Treat a limit message as a reason to inspect the dashboard. Raising an app cap is only useful when that cap is the constraint. It cannot solve an exhausted shared allowance.

Choose limits around useful work

Start with one app and one bounded task. For a new coding tool, try: “Help me investigate why this form accepts an empty email field. Inspect the relevant validation code, suggest a small fix and explain how I can check it.” Review the result before asking for more changes.

For product work, use a small, approved set of research notes: “Group these comments by user problem. Preserve the original comment IDs and list uncertainty. Do not invent frequency counts.” Check the grouping against the notes before using it in a decision.

For delivery work, try: “Turn these approved meeting notes into a draft action list. Include an owner only where one was named. Keep unresolved questions separate.” Confirm the actions with the people responsible.

These are example task briefs, not promises that every connected app supports those workflows. Choose an app that supports the task and the information you are permitted to use.

I would give an experimental app a modest starting cap, then review whether its output earns the time and allowance it uses. There is no universally correct percentage. Increase it when you have a useful reason, rather than because a tool invites you to.

Set a cap and review access

Open Settings → Usage → App limits, choose the app’s weekly percentage and save. Review the option allowing apps to use credits after included limits. To revoke the connection, use Settings → Security and login → Sign in with ChatGPT → Disconnect. OpenAI documents these controls.

Using your plan does not itself expose your conversations or memories. Review any additional permissions separately.

A usage cap controls consumption. It does not check the quality of an answer or decide whether a document is appropriate to share. Keep those decisions in your own review process.

Try a one-week review

Before testing an app, write down its purpose, starting cap and what a useful result would look like. After a few tasks, record the usage shown, corrections needed and whether you would choose the tool again.

For example: “Purpose: investigate small validation bugs. Success: I can explain the fix and reproduce the check. Review: keep, adjust the cap or disconnect.”

Use that evidence to make your next decision. A larger allowance is helpful when it supports worthwhile work. The first step is finding out which work that is.

Feature details checked on 3 November 2026. Availability and limits can change. Examples and recommendations are editorial guidance.