Claude Code Usage Limits: Weekly Caps, Credits and Monitoring

Published 23-10-2026. Last verified against Anthropic's documentation on 23-10-2026. Reviewed monthly, because plans, limits and prices change.

Quick answer

Claude Code on a Pro, Max, Team or Enterprise seat runs under two documented ceilings: a session limit that resets every five hours and a weekly limit. Anthropic publishes no fixed message or token count for either, only multipliers, so every "hours per week" figure you find elsewhere is an estimate. This page is for a developer or tech lead who runs Claude Code on a subscription seat and needs to plan a working week. Our advice: read your usage from the tool itself, cut context before you buy anything and switch on usage credits only with a spend limit set. We ran no workload to the cap, so we publish no measured number.

How we checked the information in this article

Our method is to read the vendor documentation, write down what it says with the date and mark what we did not run. On 23-10-2026 we read five Anthropic sources: the help article on using Claude Code with a Pro or Max plan, the help article on how usage and length limits work, the Claude Code cost page, the Claude Code error reference and the pricing page. Our gate for this page: a limit is stated only if Anthropic documents it or we measured it. We measured nothing, so the gate rejected what Anthropic does not document. We dropped a Portkey post that gives 40 to 80 hours of Pro usage per week, because it carries no source and was last updated on 29-07-2025. We dropped DataCamp's weekly Sonnet limit for Max, because Anthropic's help article today names a separate weekly limit for Fable instead. And we dropped DataCamp's May 2026 doubling of limits, because we found it on no Anthropic page we read. Every figure below is vendor-reported and was read on 23-10-2026.

What are the Claude Code limits on each plan?

The limits are a five-hour session cap and a weekly cap on every paid plan, expressed as multipliers rather than numbers. Anthropic's pricing page says Claude Code is included in all paid plans and is not included in Free. Pro and Max both have a five-hour session limit and a weekly limit, and Max has an extra weekly limit for Fable. Claude and Claude Code share the same limits, so a long chat in the morning leaves less for the terminal in the afternoon.

Sources: Anthropic's pricing page and its Claude Code on Pro and Max article, both read on 23-10-2026 and listed under Sources below. For what a seat costs against what Claude Code is worth, see our guide to Claude Code pricing.

What is the difference between the session cap and the weekly cap?

The session cap is a window that resets every five hours and the weekly cap is a second ceiling that resets once a week. Anthropic's help article does not state the exact reset moment. The message in the terminal does, and its format is shown in the error reference: "You've hit your session limit · resets 3:45pm" for a reset later the same day and "You've hit your weekly limit · resets Mon 12:00am" for a weekly reset.

Both caps apply to all models together, so switching with /model does not get you back to work. A message that names a model family, such as "You've hit your Opus limit", is different: switching to a model outside that family keeps you working. We would still think twice, because each model has its own prompt cache and the first request after a switch costs more. If you hit the session cap in the middle of an answer, Claude Code may keep working briefly to finish the current step and show "Usage limit reached · wrapping up". That work counts toward your weekly limit. Per Anthropic it happens at most once per weekly period on Pro and every time on Max, and it is being rolled out in stages, so your account may not have it yet.

What counts toward the limit?

All activity in Claude on the web, desktop and mobile and in Claude Code, including the IDE integrations, counts against one pool. Anthropic's help article names the length and complexity of your conversations, the features you use, the model and the effort level as the drivers, and says a higher effort level uses more tokens.

The cost page adds the reasons a long session drains the plan faster than your typing suggests. Claude Code sends the full conversation with every request. Your first message after a break longer than the cache lifetime reprocesses the whole context. A session-scoped scheduled task fires on its interval even while you are away. A subagent sends its own requests on top of the main conversation, and agent teams use roughly 7x more tokens than a standard session when teammates run in plan mode. Compacting is itself a large request, because it reads the conversation it summarizes. Idle background work is small: Anthropic puts it at typically under USD 0.04 per session.

In our view the forgotten scheduled task is the first suspect when a week empties faster than your work explains, and we would check /usage for loops before blaming the plan. Our guide to scheduled tasks and skill automation shows how to set them up so they stop when the work does, and the Claude workflow automation collection has skills for the recurring jobs themselves.

How do you check your Claude Code usage?

Run /usage inside a session. On a Pro, Max, Team or Enterprise plan it shows plan usage bars, a split of recent usage by skills, subagents, plugins and individual MCP servers as percentages, and recommendations for any behavior behind 10% or more of your recent usage. Press d or w to switch between the last 24 hours and the last 7 days. Anthropic says the figures are approximate and computed from local session history on this machine, so usage from other devices or claude.ai is not included. Anthropic's help article names /status for your remaining allocation instead. We ran neither command.

Claude Code can also warn you before the cap, with a message such as "You've used 85% of your session limit · resets 3:45pm", and you can show the same numbers continuously by adding the rate limit fields to a custom status line.

For history beyond a single machine's session, we would pick ccusage first. On 23-10-2026 its repository describes it as a tool that analyzes coding agent token usage and cost from local data, with reports by day, week, month, session and five-hour block, run with npx ccusage@latest, under the MIT license and with 19.1k stars. We put it before any paid tool because it is free and reads files you already have. What would move it down: dollar figures it computes from a price file are estimates, and on a subscription they are not a bill.

We sell skills that load into Claude, Cursor, Codex CLI and Copilot, and we name five of ours on this page. No vendor paid to be listed. To reconcile usage exports against what you were billed, LLM Usage Cost Reconciler costs USD 19.99 per its listing. LLM Cost Optimization Analyst costs USD 29.99 and only fits when you already have logs and a model-routing question. We ran neither. Prices were read from the listings on 23-10-2026. More tools for this job are in the Claude LLM ops collection.

What do you do when you hit the cap?

Wait for the reset time in the message, or use one of three documented ways to keep going. The first is auto-continue: in the Desktop app, the Code tab has an "Auto-continue when limits reset" checkbox on the session limit card, but not on the weekly limit card, and in the CLI the setting "Continue automatically at usage limit" lives in /config. The two settings are separate. The second is a model outside the family named in the message. The third is usage credits, below.

Do not mistake a usage limit for a rate limit. A usage limit means your plan reached its quota and Claude Code stops until the reset. A 429 rate limit is a short server-side throttle, and Claude Code retries it automatically, up to 10 times with growing waits, according to the error reference. We would also not read a context or auto-compact warning as a limit: it means the conversation is close to its window, not that your plan is empty.

How do usage credits work?

Usage credits let you keep working past your plan limits and bill at standard API rates, separate from your subscription. Anthropic's help article on extra usage for paid plans lists Pro, Max 5x and Max 20x as eligible. You switch them on under Settings, Usage on claude.ai, add a payment method, prepay with "Add funds" and choose a monthly spend limit or unlimited. Auto-reload is optional. Run /usage-credits in Claude Code to jump to that screen. On Team and Enterprise, a member without billing access sends a request to the admins instead. Two details catch people out. Monthly API credits that come with Max and Team plans do not apply to usage credits in Claude, Claude Code or Cowork. And credits do not change when the five-hour and weekly limits reset.

Our judgment: set a monthly spend limit on the day you enable credits. The "unlimited" choice turns a weekly cap into an open bill, and the /usage row shows no spend figure at all when no limit is set.

How do you make a week of Claude Code last longer?

Cut context first, then match the model, then trim what loads every time. This is our ranking, based on the cost page and the help articles, and not on a test of ours.

1. Clear between unrelated tasks with /clear. It sits first because every request resends the whole conversation, so stale context is paid for on every later message. Use /rename first so you can /resume the session.

2. Match model and effort to the task. Anthropic says Opus uses noticeably more quota than Sonnet and recommends Sonnet for most coding. It sits second because the saving depends on how often you really need Opus. You cannot turn thinking off on Opus 5.5, Sonnet 5.5, Haiku 5.5 or the Fable models, so lower the effort level with /effort instead.

3. Trim what loads on every request. Keep CLAUDE.md under 200 lines, move workflow-specific instructions into skills that load on demand and disable MCP servers you do not use. It sits third because it is a fixed tax per request and smaller than context growth. Our guide to reducing Claude Code token usage and the Claude prompt and skill engineering collection go through the mechanics.

4. Avoid cold-cache restarts. It sits fourth because it only bites after a break longer than the cache lifetime: compact or clear before you step away rather than returning to a huge context.

5. Keep agent teams small and give teammates Sonnet. It sits last because it only applies if you use teams, and then it moves the most: Anthropic says usage is roughly proportional to team size. Our guide to Claude Code subagents and the Claude agents orchestration collection cover the setup.

Skills that shorten replies come after all five. Tokenlean is free with an account; its listing claims about 10% fewer output tokens on average, a seller figure and not ours. Token Trimmer costs USD 7 and claims up to 70% fewer tokens without showing how that was measured. We would not buy it for the 70% alone, because Anthropic's help article on models says the conversation grows fastest, and shorter replies do not shrink the conversation. We ran neither skill. Our guide to reducing Claude Code costs with skills compares more options.

Shared Rate Budget for Multi-Agent Setups costs USD 15 and lets several agents on one account share a request budget, so a 429 seen by one agent pauses all of them. It addresses 429 throttling, not a plan's session or weekly cap, so we would not buy it for a single seat. We have not run it.

How do the limits work for a team?

On Team and Enterprise plans, each member draws from a per-seat allowance on a rolling five-hour window and a weekly window, shared with Claude chat and Cowork, and sized by the seat tier. Admins can turn on usage credits and set spend limits at the organization, group or member level, and the spend report in org analytics exports per-user spend as CSV. On the Console, spend limits sit on the workspace that Claude Code creates for you.

Anthropic's cost page gives two planning figures. Across enterprise deployments the average cost is around USD 13 per developer per active day and USD 150 to 250 per developer per month, with costs below USD 30 per active day for 90% of users. Those are Anthropic's numbers for API-billed use, not ours, and they are an average to pilot against rather than a budget. For API organizations, Anthropic recommends 200k to 300k tokens per minute and 5 to 7 requests per minute per user for a team of 1 to 5, falling to 10k to 15k tokens per minute per user above 500 users. When per-user attribution matters on Bedrock, Vertex or Foundry, Anthropic points to OpenTelemetry or an LLM gateway; our Claude DevOps and cloud collection is where we would look for the supporting skills.

Where to go next

Re-read Anthropic's own pages for anything that changes after 23-10-2026: the Claude Code on Pro and Max article, the cost page and the error reference, all listed under Sources below. This page contains no legal, tax or procurement advice, so check plan terms and invoices yourself. We will recheck it monthly and mark what changed.

Written by Samuel Rose, Founder of Agensi.

Sources

• Claude: pricing and plans
• Claude Help Center: using Claude Code with your Pro or Max plan
• GitHub: ryoppippi/ccusage, local token usage and cost reports
• Claude Code docs: costs
• Claude Code docs: error reference