Qwen Code usage on Mac

At a glance

Quota and resets

None. The free Qwen OAuth tier ended in April 2026, and the Alibaba Cloud Coding Plan and the other sign-ins show usage only in their own consoles, so there is no window and no reset.

Which account it counts against

None. Qwen Code keeps no account identity for any sign-in, so its sessions are not tied to an account.

When a turn starts and finishes

Running turns are not detected, because Qwen's records do not say when a turn ends. Completions come from Qwen's own Stop hook, which Agent HUD installs; a cancelled or failed turn reports none.

When Qwen Code is about to show its permission dialog, the request goes to the island first and Qwen waits for your answer there — ten minutes unless you choose otherwise, then Qwen shows its own dialog. Questions and plan approvals stay in Qwen. A session that was already running keeps the hooks it started with.

What it reads on your Mac

The chat transcripts in Qwen's projects folder, one usage record per model call. History copied by /branch counts once, where it happened. Calls Qwen makes for itself, such as compression and titles, are recorded nowhere and so are not counted. No credentials are read.

Common questions

Why does Qwen Code show no quota?

No sign-in Qwen Code supports reports an allowance anywhere Agent HUD can read it. Sessions and token counts are read as usual; there is simply nothing that says what is left.

Are sub-agent calls counted?

Yes. Qwen records every model call in the session that made it, sub-agent calls included, and each counts once.

Does Agent HUD read my Qwen credentials?

No. No sign-in file is read and no service is contacted; everything comes from the transcripts on your Mac.

Can I answer Qwen's permission prompts from the HUD?

Yes, with deny or allow once, for sessions started after Agent HUD installed its hook. Unanswered for ten minutes, or with Agent HUD closed, Qwen asks in its own dialog as usual.

Read next

• Quota windows and resets — Why every AI coding tool reports its own quota windows, what the percentage actually measures, and when a reset counts as real.
• In, Out and Cache: how token usage is counted — The three token dimensions Agent HUD counts, why reasoning tokens are never added twice, and why cache reads sit apart from the rest.
• Agent HUD on an external monitor — What appears at the top of a screen with no notch, how to give each display its own look, and whether a second monitor slows the Mac down.