Best for: non-urgent work
Wait for the reset shown in your account
The warning and Settings > Usage show when included usage should return. Waiting is the cleanest answer when the task is not urgent and you do not want additional metered spend.
Claude usage limit reached? See what stopped you and when it resets.
That message can mean you exhausted the current 5-hour session allowance or the separate weekly allowance on a paid plan. Waiting for the wrong reset will not help. Check the exact warning, confirm the timestamp in Settings > Usage, and keep both limits visible while you work so the next cutoff does not arrive without warning.
Quick answer
When Claude says your usage limit was reached, first open Settings > Usage and identify the limit named in the warning. The session allowance resets every five hours. Paid plans also have a separate weekly allowance that resets at a fixed day and time assigned to your account. Claude on the web, Claude Desktop, and Claude Code share the same subscription allowance, while API-key usage is billed separately. Tokens 4 Breakfast keeps the live percentages and reset times in your Mac menu bar so you can see the cutoff coming before it blocks the next task.
Know the difference
Usage limit vs context limit
These warnings can look similar, but the right fix depends on which budget ran out.
LimitWhat it meansWhat to do next
- 5-hour session
- What it meansThe short-term allowance included with your Claude plan. It is shared across Claude on the web, Claude Desktop, and Claude Code when they use the same subscription.
- What to do nextCheck the reset countdown in Settings > Usage. Wait for that reset, or use credits if your paid plan supports them.
- Weekly allowance
- What it meansA separate paid-plan cap across models. Anthropic assigns a fixed reset day and time to the account.
- What to do nextCheck the weekly reset shown in Settings > Usage. A five-hour session reset will not restore an exhausted weekly allowance.
- Context length
- What it meansThe amount of information one conversation can actively hold. It is not an account usage allowance and it can fill during a long chat or coding session.
- What to do nextStart a new chat, use Projects for reusable knowledge, or run /compact in Claude Code when the same task must continue.
- API rate or budget
- What it meansMetered API usage tied to the Claude Console or a cloud provider. It is separate from the included usage in a Claude subscription.
- What to do nextCheck the Console or cloud-provider dashboard. Review rate limits, billing limits, and API spend instead of waiting for a subscription reset.
Product proof
Your reset time, one click from the work
The Limits view keeps Claude's current session and weekly allowance visible without turning usage checking into another browser routine. The signal stays on your Mac and sits beside the other AI tools you use.
- Current 5-hour and weekly percentages
- Exact reset time and a plain-language countdown
- Claude and Codex limits in one compact view
- No T4B account, no telemetry, free for one provider
Step by step
The full walkthrough
Each step stands on its own. Skip to the one that matches where you are.
Read the exact warning before changing anything
A 5-hour session limit, a weekly allowance, a conversation-length error, and an API rate limit are different problems. The blocking message normally names the limit and shows a reset time. Capture that wording first. It tells you whether a short wait will restore access or whether the weekly allowance is the real constraint.
Check Settings > Usage for the authoritative time
On a paid Claude plan, Settings > Usage shows the current 5-hour consumption, time remaining, weekly usage, and the next weekly reset. In Claude Code, /usage shows plan usage and activity. Trust the timestamp in your account rather than a generic reset schedule copied from a forum post.
Know what the 5-hour reset restores
Claude's session-based allowance resets every five hours. The amount of useful work inside a session is not a fixed message count. Conversation length, attached files, model choice, effort level, tools, connectors, and artifacts can make one session consume the allowance much faster than another.
Check the weekly allowance separately
Paid plans have a second allowance that resets at a fixed day and time assigned to the account. Your reset schedule stays the same regardless of when you begin working. If the weekly allowance is exhausted, a 5-hour session reset will not restore it. Settings > Usage shows the exact weekly reset assigned to you.
Remember that Claude surfaces share the plan
Claude on the web, Claude Desktop, and Claude Code draw from the same subscription allowance. A heavy coding session can leave less room for chat later, and the reverse is also true. API-key usage is separate and billed by token through the Claude Console or the cloud provider attached to that key.
Reduce the usage that does not move the task forward
Start a new conversation when the job changes, keep project instructions concise, reuse recurring context through Projects, lower effort for routine work, and disable tools you do not need. These actions do not bypass the limits. They keep more of the allowance focused on finished work instead of repeated context and unnecessary tool calls.
Keep both reset clocks visible while you work
Tokens 4 Breakfast places the current Claude session percentage, weekly allowance, and reset countdown in the Mac menu bar. You can see Claude next to the rest of your AI stack, receive a warning before the limit is exhausted, and decide whether to continue, compact the task, or wait before another long run.
Pro tips
- If the message names the weekly allowance, waiting five hours will not fix it. Check the weekly reset shown in Settings > Usage.
- Treat any fixed message-count claim as an estimate. Claude usage varies with context, model, effort, files, and enabled tools.
- Use a high-effort model for the decisions that need it, then switch to a lighter workflow for routine execution.
Complete playbook
21 practical ways to make Claude usage last longer
None of these actions bypass Anthropic's limits. They reduce repeated context, unnecessary tool calls, and avoidable regeneration so more of your allowance reaches finished work.
Before a heavy Claude session
The best time to protect the allowance is before a long research, writing, or coding run begins.
- 01All Claude surfaces
Check usage before starting expensive work
Open Settings > Usage before a large task. In Claude Code, run /usage. If the session or weekly bar is already high, you can wait, narrow the job, or postpone a tool-heavy run instead of being stopped halfway through it.
- 02Diagnosis
Name the limit you are managing
Write down whether you are close to the five-hour session allowance, the weekly allowance, the conversation context window, or an API budget. Each has a different reset and a different remedy. Treating them as one limit leads to wasted waiting and unnecessary upgrades.
- 03Planning
Schedule deep work around the displayed reset
Use the reset time shown in your own account. Begin a long code review or research task after enough allowance has returned, and place lighter work near the end of the window. Generic schedules from social posts cannot tell you the reset assigned to your account.
- 04Prompting
Define the result before sending the first prompt
Include the objective, constraints, required format, source standard, and acceptance checks upfront. A precise definition of done prevents clarification loops and full regenerations caused by requirements that appeared too late.
- 05Prompting
Batch questions that use the same context
Ask for findings, risks, alternatives, and next actions in one structured request when they depend on the same material. Separate unrelated jobs, but combine closely related questions so Claude does not need several back-and-forth turns to reconstruct the same brief.
Keep conversations and Projects focused
Long histories and irrelevant files quietly make every later turn heavier. Keep only the context the current outcome needs.
- 06Claude
Start a new chat when the outcome changes
A conversation that moves from research to travel planning to code carries all three histories. If the next prompt would make sense without the current thread, start a new chat. A fresh conversation reduces unrelated context and usually improves focus as well as usage efficiency.
- 07All Claude surfaces
Repair the failed part, not the whole result
Point to the specific paragraph, function, slide, or calculation that is wrong. State what must change and what must remain untouched. Regenerating a full deliverable because one section failed spends allowance on work that was already acceptable.
- 08Projects
Put recurring source material in Projects
Store stable briefs, policies, product notes, and research sources in a Project when you will reference them repeatedly. Anthropic says reused project content is cached, so only new or uncached portions count when that knowledge is used again.
- 09Projects
Keep project instructions short and durable
Use project instructions for the lasting role, tone, rules, and quality bar. Put the current request in the chat and detailed evidence in project files. Temporary instructions left in the permanent brief create extra context and can conflict with later work.
- 10Projects
Remove stale files from active Projects
A Project should be a working knowledge base, not a permanent archive. Remove drafts, duplicate exports, and sources that no longer apply. This reduces retrieval noise and lowers the chance that Claude reads irrelevant material into the active context.
- 11Files and RAG
Scope large sources to the relevant evidence
For a large report, point Claude to the chapters, pages, rows, or sections that answer the question. Projects can use retrieval to load relevant material from a larger knowledge base. There is no official magic word limit for every file, so relevance matters more than arbitrary splitting.
Spend expensive capabilities only where they help
Model choice, effort, tools, connectors, and artifact generation all affect how quickly included usage is consumed.
- 12Templates
Reuse a stable structure for recurring work
Keep a proven outline for recurring reviews, reports, and briefs, then replace only the goal, evidence, and constraints. This reduces prompting detours, and Anthropic says similar prompts used frequently may be partially cached.
- 13Thinking
Turn off extended thinking for routine tasks
Simple extraction, formatting, short rewrites, and straightforward summaries rarely need the deepest reasoning path. Disable extended thinking for mechanical work, then restore it when complex analysis or a high-stakes decision can benefit from the extra reasoning.
- 14Effort
Lower effort when the task is straightforward
Effort level changes usage. Use a lower setting for predictable work such as classification, cleanup, or formatting, and raise it for ambiguous synthesis, architecture, or difficult debugging. The goal is deliberate routing, not permanently choosing the cheapest setting.
- 15Tools
Disable tools and connectors you do not need
Web search, Research, and MCP connectors can add substantial context and tool output. Turn them on for live retrieval or connected actions, then disable them when the source text is already available and the task only needs editing or reasoning.
- 16Models
Match the model to the difficulty of the job
Use an efficient model for lookups, mechanical edits, and well-specified execution. Reserve the highest-capability model for architecture, hard debugging, nuanced decisions, and other work where deeper reasoning can change the outcome. Check /model for the choices available to your account.
Control the Claude Code context budget
Claude Code exposes commands that show which allowance is running low and what the current session is carrying.
- 17/usage
Run /usage before a long coding run
Use /usage to inspect plan consumption before asking Claude Code to explore a repository, run a large migration, or execute a multi-file refactor. The visible bars help you decide whether to begin, reduce scope, or wait for included usage to return.
- 18/clear
Run /clear between unrelated coding tasks
/clear removes the current conversation history while keeping project instructions and files available. Use it after one task is finished and before the next unrelated job. Save any detail you still need first because clearing the chat cannot be undone.
- 19/compact
Run /compact when the same task must continue
/compact summarizes the current conversation and frees context while preserving the essential thread. It is the better choice when a long debugging or implementation task is still active and starting over would discard decisions you need.
- 20/context
Inspect context, then point to exact evidence
Run /context to see what the session has loaded. Refer to the relevant file path, function, or a short range of log lines instead of pasting an entire repository file or build log. Specific evidence keeps the session smaller and the answer more precise.
- 21/model and /cost
Use /model and /cost for the right billing path
Use /model to select an appropriate model for the next phase. If Claude Code is authenticated with an API key, /cost shows running token and dollar usage for that session. Subscription users should rely on /usage and Settings > Usage for their included allowance.
After the limit
What should you do after reaching a Claude limit?
Choose the least expensive option that keeps the work moving without turning a temporary limit into an uncontrolled bill.
Best for: urgent paid-plan work
Enable usage credits with a spending limit
Eligible paid-plan users can continue at standard pay-as-you-go rates after included usage is exhausted. Set a monthly cap before enabling credits so a long agentic session cannot create an open-ended bill.
Best for: consistently high usage
Upgrade only after cleaning up the workflow
A higher plan provides more headroom, but it does not make mixed-topic chats, repeated files, or unnecessary tools efficient. Fix those patterns first, then upgrade if clean sessions still interrupt valuable work often.
Best for: controlled API workflows
Move metered work to an API key
API-key usage is billed separately through the Claude Console or a supported cloud provider. This can keep urgent work moving, but it replaces a hard subscription stop with token-based billing. Monitor /cost and the authoritative billing dashboard.
Fact check
Common Claude limit advice that does not hold up
These claims are repeated online, but current Anthropic documentation does not support them as quota-saving rules.
Claim
Editing an earlier prompt refunds the usage
What is actually supported
Editing can create a cleaner branch, but Anthropic does not state that it reverses allowance already consumed. Use edits for clarity, not as a quota-refund mechanism.
Claim
Every uploaded file must stay below one fixed word count
What is actually supported
Anthropic publishes file and context limits, but no universal 2,000-word rule. Select the evidence the task needs and use Projects or retrieval for larger knowledge bases.
Claim
You must summarize after a fixed number of messages
What is actually supported
There is no documented 15-message or 20-message threshold. Start fresh when the task changes. In Claude Code, compact when the same task continues and context has become heavy.
Claim
Voice prompting automatically saves Claude usage
What is actually supported
Dictation may help someone give a complete brief in one pass, but a longer spoken prompt can also add more input. Anthropic does not guarantee a quota saving from voice alone.
Claim
Opening Claude on another device gives a new allowance
What is actually supported
Subscription usage is shared across Claude on the web, Claude Desktop, and Claude Code. Switching devices or surfaces does not create a separate included pool for the same account.
FAQ
Common questions
Short, direct answers to the things people ask most about this.
What does 'Claude usage limit reached' mean?
It means the account exhausted either its current 5-hour session allowance or a separate weekly allowance. Read the warning and open Settings > Usage to see which limit stopped the session and the exact time it will reset.
Does Claude reset every five hours?
The session-based usage allowance resets every five hours. Paid plans also have a separate weekly allowance, so a five-hour reset will not restore access if the weekly allowance is the limit you exhausted.
When does the Claude weekly limit reset?
It resets at a fixed day and time assigned to your account. That schedule stays the same regardless of when you start using Claude. Open Settings > Usage to see your next weekly reset.
Do Claude Code and Claude on the web share limits?
Yes when you use the same paid subscription. Claude on the web, Claude Desktop, and Claude Code count toward the same plan allowance. API-key usage is separately metered and billed.
Why did my usage run out faster than expected?
Long conversations, large files, higher effort, expensive models, Research, web search, connectors, artifacts, and agentic tool loops can all consume more of the allowance. Claude does not promise one fixed message count for every session.
Can I monitor Claude limits without keeping Settings open?
Yes. Tokens 4 Breakfast shows the current session percentage, weekly allowance, and reset times in the Mac menu bar, with optional warnings before a limit is exhausted.
Does starting a new chat reset my Claude usage limit?
No. A new chat starts with a clean conversation context, which can make later turns lighter, but it does not reset the account's five-hour or weekly allowance. Only the reset time shown in Settings > Usage restores included plan usage.
Why is my Claude usage limit not resetting?
First confirm whether the warning names the five-hour session allowance or the weekly allowance. They have different reset times. If the displayed time has passed, refresh Claude, check Anthropic's service status, and contact support if the account still shows the old state.
Is there a fixed Claude Pro message limit?
No single public message count applies to every Pro session. Anthropic says usage varies with message and file length, conversation history, model, effort, tools, connectors, and artifacts. Settings > Usage is the reliable view for your account.
Can I keep using Claude after the included limit is reached?
Eligible paid-plan users can enable usage credits and continue at pay-as-you-go rates. You can also wait for the displayed reset or use separately billed API access. Set a spending cap before turning a hard stop into metered usage.
Verified sources
Current primary documentation used for the time-sensitive details on this page.