Claude Code usage limits: 5-hour session and weekly resets explained

Updated 2026-10-25. Every limit below is quoted from an official Anthropic page and linked. API prices from models.json.

If you use Claude Code on a Pro or Max plan, you have hit, or will hit, a usage limit. This guide covers only what Anthropic officially says about those limits: how they reset, what they're shared with, and the MCP output cap. It ends with a worked example of when paying per token via the API is cheaper than a plan. Anthropic publishes no numeric token or message caps for Pro or Max, so we don't make any up.

Two limits: a 5-hour session and a weekly limit

"Paid plans have two types of usage limits: a session limit that resets every five hours, and weekly limits that reset once every week."

Source: Anthropic, Understanding usage and length limits

In short, a heavy afternoon can use up your 5-hour session allowance, which comes back when the session resets. Separately, a busy week can use up the weekly allowance, which only comes back at the weekly reset.

Pro and Max share limits between Claude and Claude Code

"Both Pro and Max plans have a five-hour session limit and a weekly limit. Max plans also have a separate weekly limit for Fable. These limits are shared across Claude and Claude Code"

Source: Anthropic, Using Claude Code with your Pro or Max plan

So chatting in the Claude app and coding in Claude Code draw from the same allowance. On Max, Fable usage also counts against its own weekly limit.

Plan prices and what "more usage" means

Claude Pro: $20/month billed monthly, or $17/month billed annually ($200 up front). Includes Claude Code. (claude.com/pricing)
Claude Max: the pricing page says "Choose 5x or 20x more usage than Pro*", with no absolute numbers. Max 5x is $100/month and Max 20x is $200/month. (pricing, What is the Max plan?)
After a limit: Anthropic's pricing page says paid plans can keep going with usage credits billed at standard API rates. (pricing)
No published caps: neither page gives a token or message count for a session or a week, so nobody outside Anthropic can say exactly how many tasks a plan covers.

To see where your usage goes, Claude Code's official changelog lists usage attribution in the /usage dialog "showing cache misses, long context, subagents, and per-skill/agent/plugin/MCP breakdowns over the last 24h or 7d". (Claude Code CHANGELOG)

MCP tool output limits in Claude Code

"Claude Code displays a warning when MCP tool output exceeds 10,000 tokens and limits output to 25,000 tokens by default. To change the token limit, set the MAX_MCP_OUTPUT_TOKENS environment variable, for example MAX_MCP_OUTPUT_TOKENS=50000. The warning threshold is fixed."

The variable "applies to tools that don't declare their own limit. Tools that set anthropic/maxResultSizeChars use that value instead for text content".

Source: Claude Code docs, MCP

Raising the cap lets bigger tool results into the context, and every token of tool output is context that uses up your allowance (or your API bill). Tool schemas cost tokens too. See how much MCP tools cost in tokens.

When paying per token via the API beats a plan

A plan is a flat price with limits you can't see in numbers. The API has no plan limits, but you pay for every token. The API is cheaper whenever your monthly API cost is below the plan price. Break-even tasks per month = plan price / API cost per task.

Worked example (an illustration, not a measurement): one task = 100,000 input tokens (90% cache hits) and 8,000 output tokens, over 22 working days a month. Prices per 1M tokens are input / cached input / output.

Model	$ per 1M (in / cached / out)	API cost per task	5 tasks/day	20 tasks/day	50 tasks/day	Break-even vs Pro ($20)	vs Max 5x ($100)	vs Max 20x ($200)
Claude Opus 5.5	$4 / $0.2 / $20	$0.218	$23.98	$95.92	$239.80	92	459	917
Claude Sonnet 5.5	$2 / $0.1 / $10	$0.109	$11.99	$47.96	$119.90	183	917	1,835

Example figures computed from models.json (standard tier, last updated 2026-10-25). Cache-write fees are left out, so real API costs can be a little higher. Break-even columns are tasks per month. Above the break-even a plan is cheaper only if your usage fits inside its 5-hour and weekly limits, which Anthropic doesn't publish as numbers.

The API tends to win for light or bursty use, below the break-even, and when you'd otherwise hit plan limits and wait for a reset.
A plan tends to win for steady daily use well above the break-even that stays within the limits.

Plug in your own tasks, tokens and model in the TokenTab calculator or the subscription vs API calculator, and see Claude Code cost per month for light, medium and heavy usage.

FAQ

How often do Claude Code usage limits reset?

Anthropic says paid plans have two types of usage limits: a session limit that resets every five hours, and weekly limits that reset once every week.

Are Claude Code limits separate from Claude chat limits?

No. Anthropic says Pro and Max plans have a five-hour session limit and a weekly limit, and that these limits are shared across Claude and Claude Code.

How many messages or tokens does a Claude Pro or Max plan include?

Anthropic does not publish numeric caps. Max is described as 5x or 20x Pro's usage per session, but no absolute token or message counts are given, so TokenTab does not estimate them.

What is the MCP tool output limit in Claude Code?

Claude Code shows a warning when MCP tool output exceeds 10,000 tokens and limits output to 25,000 tokens by default. You can raise the limit with the MAX_MCP_OUTPUT_TOKENS environment variable, for example MAX_MCP_OUTPUT_TOKENS=50000. The warning threshold is fixed.

When is the API cheaper than a Claude plan?

When your monthly API cost is below the plan price. Divide the plan price by your API cost per task to get the break-even tasks per month. Above that, the plan is cheaper only if your usage fits inside its session and weekly limits.

Run your own numbers

Open the TokenTab calculator Claude Code cost per month MCP tool token cost