Agent HUD · Guides · Supported clients
GLM coding plan quota on Mac
GLM has no logs of its own here: the quota comes from the service, and the tokens come from whichever client used it.
At a glance
| Quota | Token, credit and time limits |
|---|---|
| Account | Billing pool, per credential |
| Running · completion | From the calling client |
| Credentials read | Plan keys only |
What you see
The limits the service lists — token, credit and time limits — each with its own period and reset time, and the plan name from the same response. The time limit is the MCP window and never stands in for model quota.
How the account is named
A billing pool. Because GLM offers no way to tell whether two keys belong to one account, different keys stay different pools.
Running turns and completions
Not applicable. GLM records nothing locally, so running turns and completions come from the client that made the request.
What Agent HUD reads
Nothing of GLM's own. Tokens are attributed from the OpenCode, Pi or Claude Code logs that used the service, under the provider they were recorded with.
Common questions
- Why does GLM show quota but no sessions of its own?
- Because GLM is a service rather than a client. The sessions belong to whatever ran the request — OpenCode, Pi or Claude Code — and the quota belongs to the plan behind the key.
- I have two GLM keys. Why are they not combined?
- There is no way to check whether two keys belong to one account, and guessing would merge quotas that are actually separate. Each key stays its own pool.
Last updated: 2026-09-20 · The behaviour described here is documented in full, with the code it comes from, in Agent HUD Open. providers.md