Start with the account’s current usage display
Open /usage inside Claude Code and check the account’s usage settings when a run is interrupted. Read the affected allowance and any reset information shown for your account. A third-party countdown or an old message-count estimate cannot tell you what your next task will cost.
The first question is what actually stopped the work: subscription capacity, an API limit, a long conversation, a permission prompt, or an unavailable service. Those require different next steps. Copy the exact error into your task notes before changing settings or starting replacement sessions.
Read /usage without confusing usage and billing
/usageAnthropic documents plan usage bars for subscribers and session token statistics on the same screen. The local dollar estimate in the session block is not a subscription invoice. API users should use the provider’s billing console as the authoritative spend record. If usage information is marked as last-known, treat it as stale and retry before planning a long run.
Keep three details together in your own log: which account the session uses, when you checked the meter, and which allowance is constrained. This avoids comparing a work organization’s usage with a personal subscription, or a fresh reading with one from before several parallel runs.
Identify the kind of limit before responding
| What you see | What it means for your next step |
|---|---|
| A plan allowance and reset time | Use the account’s displayed reset information. Starting another session on that account does not create a separate allowance. |
| A context or conversation-length warning | Reduce irrelevant context or start a fresh task with a written handoff. That is a conversation-management problem. |
| An API rate-limit response | Inspect the API account’s current rate and spend limits; reducing concurrent requests may be necessary. |
| A permission question or tool denial | Resolve the specific permission or scope issue. It is not evidence that the subscription is exhausted. |
| Usage cannot be loaded | Check the data age and retry. Missing telemetry is not a reading of zero consumption. |
| A service or authentication error | Check the provider’s status and the session’s account before treating it as a capacity problem. |
When a run must pause, preserve the changed files and a short handoff: what is done, the current branch, commands already run, and the next unresolved step. That lets you resume or switch tools without asking another agent to rediscover the whole task.
Why there is no fixed number of coding prompts
Anthropic’s help center distinguishes usage limits from conversation-length limits and explains that consumption depends on factors such as the model and conversation size. A short question and a repository task with repeated tool calls are different workloads. Plan terms can also change, so this guide does not turn a subscription into a promised number of coding tasks.
For your own planning, group work by outcome rather than prompt count. A small test repair, a broad investigation, and a long refactor should have separate observations. Record whether the result was accepted and how much human correction it required; a run that consumed little but needed a full rewrite was not necessarily efficient.
Use an interruption log before changing plans
Record interruptions for a normal working week, then another if the first was unusual. This is a decision aid, not a precise forecast of the next reset window. Use the provider’s actual numbers when available, and leave a field blank when you cannot verify it.
Checked at:
Account / provider:
Model and task:
Allowance or error shown:
Provider reset time, if shown:
Time genuinely blocked:
Alternative useful work completed:
Accepted result / remaining work:- Count time actually lost to the interruption, excluding time you would have spent reviewing or doing other work.
- Check whether a smaller task or clearer acceptance criteria avoids repeating the same investigation.
- Separate recurring capacity problems from one unusually large task or a service incident.
- Compare current plan terms only after you can describe the recurring work that extra capacity would recover.
For that last decision, use the Max vs Pro guide alongside current Claude pricing. Enabling separately billed usage is a spending decision; review the account’s current rates and controls first.
What CmdBrief’s usage view can and cannot tell you
CmdBrief shows supported provider usage data alongside agent workspaces. Claude and Codex can expose short and weekly windows; other providers expose different measurements, and some do not expose a usable meter. An absent meter is not a claim of unlimited access.
In CmdBrief 0.5.0, pace estimates distinguish recent activity from the whole-window average and stop projecting from stale measurements. A forecast still depends on past activity. A new task, model, or burst of parallel work can change the pace, so use it to decide what to inspect rather than as a promise of remaining tasks.
CmdBrief cannot raise a provider’s allowance, change its reset, or determine your final invoice. Resolve account-level questions in the provider’s own interface. Its practical role is to keep the available usage signal and the agent needing attention in the same place. See the current release notes for the app changes.
Questions
How do I check Claude Code usage?
Run /usage inside an interactive Claude Code session, then check the account’s usage settings for the current allowance and billing context. Inspect the data age if the reading is marked as last-known.
When does my Claude Code limit reset?
Use the reset information displayed for the affected allowance in your account. Do not infer it from a fixed local countdown or a reset time quoted for someone else’s plan.
Does opening a new session reset my usage?
No. A fresh conversation does not create a new provider allowance for the same account.
Can CmdBrief increase my Claude Code limit?
No. It displays supported provider data and helps you plan work. Capacity and billing remain controlled by the provider.
Still unsure? Write to [email protected] or use the contact form. Contact CmdBrief →
Primary sources
Provider terms and product behavior can change. Confirm plan, pricing, model, and permission details in the current official documentation.