Claude Code Usage Limits Explained: What "Limit Reached" Means and What to Do
Hit "usage limit reached" in Claude Code? How the five-hour session window and weekly limits work on Pro and Max, how to check what you've used with /usage, and practical ways to make your allowance last.
You're in the middle of a task and Claude Code stops with a message saying you've reached your usage limit, with a time when it resets. It's one of the most common surprises for new subscribers. Here's how the limits actually work, how to see where you stand, and how to get more done before you hit them.
Limits and plan details below are as published by Anthropic in October 2026. Anthropic adjusts them from time to time — the pricing page is the source of truth.
Two kinds of limit
If you use Claude Code with a Pro or Max subscription, your usage is measured in two windows:
- A session limit that resets five hours after your first message. Start working at 9:00 and that window resets at 14:00, however much or little you used.
- A weekly limit across all your usage. It exists so that a few people running agents around the clock don't use capacity everyone else needs.
On top of those, there can be model-specific limits — a message like "you've hit your Opus limit" — which leave other models available.
Your Claude Code usage and your ordinary Claude chat usage (claude.ai, the desktop and mobile apps) share the same allowance. A long research conversation in the morning means less coding in the afternoon.
How big are the limits?
Anthropic doesn't publish them as a fixed number of messages or tokens, because a "message" can cost wildly different amounts. A short question about one file is cheap. A request that makes the agent read 40 files, run the test suite five times and rewrite a module is expensive.
As a rough guide to the relative sizes:
| Plan | Price | Usage relative to Pro |
|---|---|---|
| Pro | $20/month | 1× |
| Max 5× | $100/month | about 5× per session |
| Max 20× | $200/month | about 20× per session |
How to check your usage
Run this inside Claude Code:
/usage
It shows how much of your current session and weekly allowance you've used, and when each resets. Checking it before starting a big task is the simplest habit that prevents mid-task stops.
What happens when you hit the limit
Claude Code stops and tells you when the window resets. Nothing is lost: your conversation, your files and any work the agent already did stay exactly where they are. When the window resets you can resume the session and carry on.
Your options in the meantime:
- Wait for the reset time shown.
- Switch model with
/modelif you've hit a model-specific limit — see how to change the model. - Upgrade from Pro to Max, or Max 5× to 20×.
- Use an API key for overflow. You pay per token instead; there's no five-hour window, only rate limits and your budget. Some plans also let you buy extra usage at API rates.
Why you hit the limit faster than expected
Usage is driven by tokens — the text the model reads and writes — not by how many messages you send. (See what are tokens in AI?.) The things that burn through an allowance fastest:
- Very long conversations. Every turn re-sends the conversation so far. A two-hour session on one thread gets expensive per message.
- Big files and big command output. Asking the agent to read a 5,000-line file or dumping a full test log into the conversation costs a lot.
- Using the most capable model for everything. Opus is more expensive per token than Sonnet; Haiku is cheaper still.
- Retry loops. An agent stuck fixing the same failing test ten times uses ten times the tokens. See stuck in an AI fix loop?.
- Parallel agents. Three agents running at once use roughly three times as fast.
How to make your limits last
- Start fresh conversations for new tasks with
/clear, and use/compactwhen a long task needs to keep going. See /compact vs /clear. - Use Sonnet for routine work and save Opus for hard problems.
- Lower effort for simple tasks with
/effort. - Keep a good
CLAUDE.mdso the agent doesn't rediscover your project every session. See how to write a CLAUDE.md. - Point the agent at specific files instead of "look around the codebase".
- Plan first with plan mode, so the expensive implementation happens once.
- Do the heaviest work early in a fresh five-hour window rather than at the end of one.
More in how to keep Claude Code costs down.
The summary
- Pro and Max have a five-hour session limit and a weekly limit, shared with Claude chat.
/usageshows where you stand and when things reset.- Hitting a limit pauses work; it doesn't lose it.
- Usage is driven by tokens: long conversations, big outputs and expensive models use it fastest.
- Clear often, pick the right model, and plan before implementing.
EasySpawn hosts Claude Code on a persistent server, so when a limit pauses your work, the session, files and running app are all still there when it resets — from any device. See how it works or join the waitlist.
Related: Claude Pro vs Max vs API Key · How to Change the Model in Claude Code · Claude Opus vs Sonnet vs Haiku · What Is a Context Window?
Keep reading
Claude Opus vs Sonnet vs Haiku: Which Model Should You Use?
Anthropic's Claude models come in tiers — Opus, Sonnet and Haiku, plus Fable at the top. What each is good at, what they cost on the API, and a simple rule for choosing for coding, chat and features inside your own app.
How to Change the Model in Claude Code (and Which One to Pick)
Switch models in Claude Code with /model, the --model flag, an environment variable or settings.json. What the aliases mean — opus, sonnet, haiku, fable, opusplan, [1m] — which is the default, and how to choose for cost and quality.