Claude does not have one public “messages per day” limit. If the app stops you, copy the exact warning and open Settings > Usage in the same account. That page—not a third-party quota chart—tells you whether you reached a five-hour session allowance, a weekly counter, or a paid usage-credit state and when that specific counter resets.
Use this first-screen decoder:
| What you see | Limit owner | What actually helps |
|---|---|---|
Approaching 5-hour limit | Included session allowance | Finish the highest-value work first; read the displayed time remaining |
5-hour limit reached - resets [time] | Included session allowance | Wait for the shown reset, change plan, or use paid credits if eligible |
| Weekly counter in Settings > Usage | Plan weekly allowance | Use its account-assigned reset time; a new five-hour window will not refill it |
continuing with usage credits | Metered paid overflow | Check balance, monthly cap, and auto-reload before continuing |
| Conversation/context warning | Length of one chat | Start a clean chat or manage context; waiting for a quota reset is the wrong fix |
| API 429 or Console limit | API organization | Check Console rate/spend limits; a Pro or Max subscription does not fix it |
This distinction matters because “daily limit” is only search vocabulary. Claude's current consumer contract is a set of different counters, not one bucket that refreshes at midnight.
What the five-hour reset means—and what it does not
Anthropic's current getting-started guidance describes Free as a session-based limit that resets every five hours. Available messages vary with demand, and Anthropic can apply other fair-access limits. There is no official fixed Free message or token count to multiply into a daily total.
The current Pro plan page says Pro offers at least five times Free usage per session during peak periods. Its included session allowance also resets every five hours. Pro additionally has a weekly all-model limit with a fixed reset day and time assigned to the account.
Max expands session headroom rather than removing limits. The Max plan page describes Max 5x and Max 20x relative to Pro, plus an all-model weekly counter and another Sonnet-only weekly counter. The exact weekly reset is visible in Settings > Usage.
None of these pages says every message independently “ages out” exactly five hours after it was sent, or that everyone resets at midnight. Use the timer Claude shows your account. If the timer says the session resets at 3:20 PM, that is the operational answer for that session.
The weekly counter can stop you after a session resets
A common surprise is returning after five hours and still being constrained. That can happen because the short session allowance and the weekly allowance are different owners.
For Pro, the public plan page describes one all-model weekly limit. For Max, it describes all-model plus Sonnet-only weekly limits. Team and legacy seat-based Enterprise contracts can expose their own seat-level counters.
There is also a live documentation conflict worth stating honestly. Anthropic's current usage-best-practices page refers to weekly counters as “Opus only” and “all other models,” while current Max and Team plan pages use “all models” and “Sonnet only.” Do not infer that every account has the same labels. Read the labels and timers shown in your signed-in Usage page; support copy and packaging can change.
Starting a new conversation does not refill a weekly allowance. Switching a model helps only if the Usage page shows a separate model-specific bucket with remaining capacity and the task can genuinely use that model.
Why one prompt can consume far more than another
Message count is a poor unit because Claude processes workloads, not identical chat bubbles. Anthropic's usage and length explanation names conversation length and complexity, features, model, and effort as consumption factors.
The practical high-burn patterns are long accumulated chats, large attachments, Research or web work, connectors and tools, high effort, and agentic or coding tasks. Project content can benefit from caching, so it is also wrong to assume that every reused project file is charged from scratch in full.
You can reduce future burn by starting a clean chat for a new topic, removing unnecessary files or tools, lowering effort for routine work, batching a clear request, and choosing a suitable model. These actions do not refund included usage you already spent.
Claude, Desktop, and Claude Code share the consumer pool
For a paid consumer account, activity across claude.ai, Claude Desktop, and subscription-authenticated Claude Code counts toward the same included usage. A heavy terminal session can reduce what remains in web chat, and vice versa.
Authentication changes the contract. If Claude Code uses an ANTHROPIC_API_KEY, its tokens are billed to the API account rather than the consumer subscription. Detailed terminal diagnosis belongs in our Claude Code usage guide; a rate limit reached workflow is covered by the Claude Code limit-reached guide.
One current edge case also deserves a stop sign: Anthropic announced a June 15 change for Agent SDK, claude -p, and third-party app usage, then paused it. The announced monthly Agent SDK credit is not currently available, and those surfaces still draw from subscription limits for now. Do not plan capacity from the superseded future-state table.
Usage credits continue work, but they create a separate charge
Paid Pro and Max subscribers can enable usage credits. After included usage is exhausted, funded credits can continue Claude and Claude Code work at standard API rates.
Credits do not enlarge or reschedule the included five-hour window. They are prepaid, metered overflow charged separately from the subscription. Configure them on the web under Settings > Usage, set a conservative monthly cap, review auto-reload, and watch the credit balance. Mobile subscribers still configure and buy credits on the web.
Anthropic also offers discounted usage bundles on eligible plans. Those are another paid overflow purchase, not included “daily allowance.” If you need the cost decision, use our Claude extra-usage cost guide rather than turning this page into a price table.
A context warning is not a timed quota
Usage limits control quantity over time. Context or length limits control how much information one conversation can process. Current paid context sizes vary by model and surface, with current support documentation describing 1M-, 500K-, and 200K-class windows. The old shortcut “paid Claude is 200K” is no longer generally correct.
If one chat is too long, create a new conversation, use Projects or retrieval, and remove irrelevant material. Waiting five hours does not shrink the conversation. Conversely, starting a new chat can reduce future token use but does not reset the account's session or weekly allowance.
Enterprise readers must identify their contract
Current usage-based Enterprise is not simply a larger included daily bucket. The access seat and usage charges are separate, usage is billed at standard API rates, and administrators set spend controls. Older seat-based Enterprise contracts can still have included allowances and usage credits until migration or renewal.
Ask the administrator which contract the organization has and look at the available Usage/analytics page. A universal “Enterprise resets every five hours” statement would be wrong.
The safe next action
- Five-hour included limit: use the displayed reset time, wait, or choose a permitted paid route.
- Weekly limit: plan around the fixed account reset; do not expect a new session to refill it.
- Usage credits: continue only after checking the separate cost cap and auto-reload.
- Context limit: reduce or restart the conversation; do not wait for a timer.
- API limit: inspect Console rate and spend controls; subscriptions and API billing are separate. See our Claude API access guide for that boundary.
The durable answer to “What is Claude's daily limit?” is therefore: read the exact account counter first. Its owner, reset, and billing consequence determine the fix; a guessed messages-per-day number does not.
