Claude usage limit reached? Recover the work before the reset.
If Claude usage limit, Claude Code usage limit, Claude Pro usage limit, or Claude Code limit reset searches brought you here, the first job is not to guess the quota math. Preserve the coding state, identify the limit type, reduce verified context pressure, and continue safely.
git status --short, git diff --stat, changed files, failed attempts, test state, and the next safe command before clearing or abandoning the session.Search phrases this guide covers.
People describe this same interruption several ways: claude usage limit, claude usage limit reached, claude code usage limit, claude code limit, claude pro usage limit, claude weekly limit, and when does claude limit reset. These phrases are not always the same technical condition, so separate the limit type before clearing context or switching tools.
Claude usage limit
Usually means account or plan allowance, shared across Claude surfaces. Confirm it in the product UI or with /usage in Claude Code.
Claude Code usage limit
Often appears during a terminal or IDE coding session. The local task may stop even though your repo still contains useful partial work.
Claude code limit reset
The reset time depends on the limit shown to your account. Preserve the task before waiting so the next session starts from verified state.
Claude Pro usage limit
Pro usage can include session-based limits and weekly limits. Claude Code activity can count against the same subscription allowance.
Three different problems can look like one limit.
Normal long-session growth
Claude's official guidance says conversation history grows fastest. File reads, tool results, model choice, effort, and long debugging threads all increase usage.
Shared account usage
Claude, Claude Code, and Claude Desktop can draw from the same plan allowance. The local terminal is not necessarily the complete account-level picture.
Usage does not match activity
Public Claude Code issue reports document cases where the account shows 100% while visible local usage does not explain it. That is evidence of a mismatch, not proof of a rogue process.
Which Claude limit did you hit?
Session reset window
Claude Pro documentation says session-based usage can reset every five hours, while account-specific weekly reset times may also apply. Treat the visible reset time as source of truth for your account.
Weekly allowance
Weekly usage limits can apply across models and reset at the fixed day and time assigned to your account. Check Settings > Usage when available.
Claude Code plan usage
Claude Code signed in through Pro or Max shares usage with Claude on web, desktop, mobile, terminal, and supported IDEs.
API key billing
If Claude Code is using an API key, the behavior is different from subscription usage. Use /cost and the Console or cloud dashboard instead of assuming a Pro or Max quota.
Diagnose before you delete anything.
/usage and save the current-session and weekly percentages, reset time, model, and plan. If using API billing, use /cost for the local session as a separate measure./context. A large conversation, many tool definitions, broad instructions, and repeated file reads can explain fast consumption even when the number of prompts looks small./usage output. Update Claude Code with claude update and report the discrepancy to Anthropic rather than deleting credentials or caches blindly./compact when the current task must continue. Use /clear only after preserving the task when the old context is no longer useful. Switch model or effort level when the task does not need the most expensive option.Normal fast usage versus an abnormal mismatch.
Likely normal
The context indicator is high, the session has many file reads or tool calls, Opus or high effort is active, and usage rises while work is visibly running.
Needs investigation
Usage jumps after one trivial prompt, the account is exhausted immediately after reset, or the local transcript and /usage summary do not explain the account-level percentage.
Possible account concern
Usage rises while all known sessions are inactive. Review signed-in surfaces, active credentials, and account security, then contact Anthropic with timestamps. Do not publish a hidden-process diagnosis without evidence.
Fast recovery checklist.
/compact for a continuation or /clear only when starting a genuinely fresh task.Recover the task before the reset.
Goal:
Current branch and changed files:
What is verified on disk:
Exact usage-limit message and reset time:
Last successful action:
Failed attempts:
Tests passed, failed, or not run:
Next safe command:After access returns, start with the verification command. Do not ask the next session to continue blindly from a transcript that stopped during a tool action.
Query-to-action map.
claude usage limit reached
Preserve the task and check whether the visible limit is session, weekly, shared-product, or API billing related.
claude code usage limit
Use /usage, /context, and the repo diff before ending the terminal session.
when does claude limit reset
Use the reset time shown in your account or limit message. Do not rely on generic reset claims when your account displays a specific time.
claude pro usage limit
Check the shared allowance across Claude and Claude Code, then decide whether to wait, enable usage credits, upgrade, or switch tools for the current task.
How ShardStitch helps with Claude Code.
ShardStitch can read persisted Claude Code session files locally and combine the useful conversation state with the repository's current facts. It creates a trust-labeled continuation packet for a fresh Claude Code session or another supported coding tool.
Conversation context
Recover persisted user and assistant messages from the selected Claude Code session when those local session files remain available.
Disk truth
Combine the conversation with changed files, Git diff, recent commits, local notes, and verification results.
Hard limit
ShardStitch cannot increase the allowance, fix Anthropic accounting, or attribute remote usage that is not represented locally.
Sources and verification.
Anthropic documents that history, model choice, and tool-heavy work affect usage; that Claude surfaces can share the same allowance; and that /compact, /clear, /context, /usage, and model selection are the supported controls. The Claude Code issue tracker also contains detailed first-party reports of account percentages that do not match visible local usage.
- Claude Help Center: models, usage, and limits in Claude Code
- Claude Help Center: usage and length limits
- Claude Help Center: use Claude Code with Pro or Max
- Claude Help Center: Pro plan usage limits
- Claude Code cheatsheet: /usage, /context, /compact, /clear
- Claude Code issue: local usage does not explain 100% account usage
- Claude Code issue: recovering context after a usage limit
FAQ.
Why is Claude Code usage draining so fast?
Long context, tool-heavy work, model choice, effort, and shared usage from other Claude surfaces can all contribute. If visible activity does not match the account percentage, preserve the evidence and treat it as an attribution issue.
Should I delete the ~/.claude directory?
No. That directory can contain valuable persisted sessions and configuration. Preserve the session and use supported diagnostics before deleting local data.
Can I continue in another tool while waiting?
Yes. A verified continuation packet can carry the goal, changed files, known failures, and next test into another supported tool without pretending the limit itself was fixed.
Does ShardStitch bypass Claude limits?
No. It recovers the coding state around the interruption; it does not change Anthropic's quota or billing.
Scope note: Usage policy and accounting can change. This page distinguishes documented behavior from reported mismatches and does not diagnose an unverified local process.