Claude’s limits show up in the middle of your work, never at the start. In the morning, Claude responds quickly, and in the afternoon, it stops you halfway through a spreadsheet. Anthropic doesn’t say how many messages fit into the limit, because that depends on what you’re doing and how. So nobody knows in advance, and the stop comes when it suits you least.
I know this from my own work. I use Claude every day, often in several conversations at once, and they all draw from one limit. I see the same thing with clients. If you do a lot of work every day on the cheapest paid plan, you’ll hit the limit sooner than you’d expect.
The limit measures the work done
Claude’s paid plans measure the work done in two windows. According to Anthropic’s help center as of October 11, 2026, one resets every five hours and the other once a week. The web app, the desktop app, and Claude Code (the version of Claude that works directly with files on your computer) all draw from both windows. According to the help center, everything you do with Claude counts toward the limit. Whatever you use up in Claude Code is then missing in chat too.
Weekly limit
The second limit window on Claude’s paid plans. It resets on a fixed day and hour assigned to your account, regardless of when you started working. Once you use it up, you wait for your reset day, even if the five-hour window resets in the meantime, unless you turn on usage credits or move to a higher plan.
A pricier plan is only the second step. The first is knowing what eats into the limit and starting big jobs in a fresh window.
What eats into the limit the most
Long tasks and long conversations use up the most. From the list in Anthropic’s help center, these are the items that come up most in everyday work:
| What uses up the limit | How to cut usage |
|---|---|
| A long conversation that Claude has to summarize | a new conversation with a summary of where you left off |
| Large attachments sent over and over | upload the materials to a project once |
| Web search, the Research feature, and connectors | use them only on tasks that need them |
| The Opus model for routine work in Claude Code | Sonnet, as the help center recommends |
| A high effort level | the default level, higher only with a reason |
| Multi-step tasks, several agents at once | a fresh window, and an eye on the agent toggles |
Connectors (links between Claude and your email, calendar, or other apps) are token-intensive, according to the help center. Tokens are the small chunks of text that the work is measured in. Connectors are part of the wrapper around the model that you don’t see on screen but that still counts toward the limit. A higher effort level works the same way: Claude thinks longer, and you burn through the limit sooner.
A switched-on toggle for running several agents at once drains it just as fast: even a small task then gets swarmed by a whole group of agents, meaning Claudes that split the work among themselves.
Where to see how much you have left
You’ll find your limit status in Settings › Usage. The Current session meter shows how much of the five-hour window you’ve used and how long until it resets. The Weekly limits meter shows your weekly usage and the reset day, and on the Max plan, it shows Fable models separately too.
As of October 11, 2026, both meters are available on the Pro, Max, Team, and seat-based Enterprise plans. According to the help center, Claude Code warns you before you hit the limit. Occasionally, Anthropic grants a free reset, and a Reset for free button appears in Usage. It restores the five-hour or weekly limit in full, but you’ll find it only on the web and in the desktop app.
Before you start a big task
Check both meters in Settings › Usage. If there’s little left in the five-hour window and it resets in half an hour, wait and start the task in the fresh window. You’ll save yourself from picking up work that the limit cut off halfway through.
How to spread out your work so the limit lasts
Start big tasks in a fresh window and leave small stuff for the end of it. You can plan around the five-hour rhythm like a schedule. End a conversation that has turned into an endless thread. Ask for a summary of where you left off and continue in a new one. Materials you keep sending over and over belong in a project (a dedicated space for one area of work), uploaded once. According to the help center, a project’s reused content counts less toward the limit, but the savings expire after a period of inactivity.
The weekly window needs a different schedule: plan your biggest batches for the days right after it resets. Long recurring tasks that aren’t urgent can be scheduled overnight in Cowork (the version of Claude you hand an entire job to). They run in the cloud even with your computer off, but they still draw from the weekly limit. And every new attempt at the same task draws from the limit again, so a precise brief saves limit too.
When you run out anyway
Once you’ve used it up, the help center gives you three options: wait for the reset, move to a higher plan, or turn on usage credits and keep going for an extra fee.
Usage credits (paid extra usage)
A way to keep going after you use up the limit on the Pro, Max, and Team plans. Your work doesn’t stop. From that point on, you just pay for usage at API prices (the developer access billed per token). It’s billed separately, outside your subscription.
You turn them on in Settings › Usage, in the Usage credits section, with the Enable button. You enter a card and set a monthly cap with Adjust limit. I’d leave the Set to unlimited option alone: without a cap, a few demanding afternoons turn into a line item that throws off your entire AI bill.
If you use them regularly, as of October 11, 2026 you can prepay for them in discounted bundles: $50 of credit for $45, $250 for $200, and $1,000 for $700. Don’t confuse them with the API credits that Max and Team plans get. According to the help center, those don’t apply to usage credits.
A higher plan or a second account?
A higher plan makes sense when you hit the limit regularly, several times a week, and waiting slows your work down. According to the price list as of October 11, 2026, Pro costs $20 a month. Max 5x at $100 gives you five times the Pro limit per window, and Max 20x at $200 gives you twenty times. Both are billed monthly only. These are US prices before tax, and the price in the Czech Republic may differ.
A second account is a second subscription for the same person. One online community owner runs a second account alongside her main one so work doesn’t stall when one of them hits the limit. But then projects, history, and settings are split across two places. And under Anthropic’s Consumer Terms, you can’t give anyone else access to your account, so colleagues need their own seats on the Team plan.
If you use Claude to rewrite an email now and then, you don’t need to worry about limits. If you hit the limit once a month, grab a coffee and wait for the new window. Upgrading because of one afternoon is expensive insurance.
The whole thing in a nutshell
- Paid plans have a five-hour window and a weekly window, shared by every version of Claude.
- Long conversations, connectors, high effort levels, and multi-step tasks eat the most.
- Keep an eye on Settings › Usage, especially before a big task.
- Start big work in a fresh window, and put your materials in a project.
- When you run out, wait, or turn on usage credits with a monthly cap.
- Buy a higher plan only once you’re hitting the limit regularly.
All of this is accurate as of October 11, 2026.
So that afternoon stop halfway through a spreadsheet is usually more about timing than about your plan. If you want your projects, model, effort level, and agent toggles set up so the limit lasts, we’ll go through them together in a 1:1 Claude setup on your computer. After that, all you need is a quick look at Settings › Usage now and then, like checking the fuel gauge before a long drive.
