Claude Code weekly limit explained: reset, plans and what drains it

What is the Claude Code weekly limit?

Paid plans have two allowances. The Help Center says they have “a session limit that resets every five hours” and “weekly limits that reset once every week” (usage and length limits). For Pro and Max, the Claude Code page describes a five-hour session limit and a weekly limit, shared between Claude and Claude Code (Using Claude Code with your Pro or Max plan).

The two run together. Anthropic's error reference says usage counts against the session and the weekly allowance at the same time, and that one burst of heavy activity, such as a large workflow fanout, can exhaust the weekly allowance before the session window resets (usage limit errors).

How does the weekly limit reset?

The Help Center says weekly limits “reset once every week” and that the Usage page shows when yours resets (usage limit best practices). The same page does not say when the five-hour session starts or how the weekly reset is scheduled beyond what the Usage page displays. Claude Code's error messages carry the time. The reference shows messages of the form You've hit your weekly limit · resets Mon 12:00am, which is an example of the format, not a promise that your reset is on Monday (errors).

When you hit a limit, Claude Code blocks further requests until the reset time in the message. Switching models with /model does not help for the session or weekly limit, because both are shared across models. In an interactive session on a claude.ai subscription, Claude Code can wait and continue the interrupted task after the reset (v2.1.234 or later), and /usage-credits manages extra usage on Pro and Max.

How do Pro and Max differ on the weekly limit?

Anthropic says plans “have different usage allowances, with paid plans offering higher limits” and that usage also depends on conversation length, features, model and effort (Help Center). The Pro and Max page names Max 5x and Max 20x but gives no message counts or token quotas (Help Center).

So there is no honest table of “Pro = N, Max 5x = M” to publish. Any page that prints one is estimating or quoting an older announcement. If a number matters to you, read your own Usage page and your own /usage.

How do I see my weekly usage?

• Settings > Usage on claude.ai: progress bars for the session and the weekly limit, plus the time remaining (Help Center).
• /usage in Claude Code: on Pro, Max, Team and Enterprise it shows plan usage bars and a breakdown of recent usage by skills, subagents, plugins and MCP servers, switchable between the last 24 hours and 7 days. It is computed from this machine's local history, so other devices and claude.ai are not included (costs).
• Status line: add the rate_limits fields to a custom status line to watch the allowance continuously (status line docs).
• Warnings: Claude Code can warn you before a window runs out, with a message such as “You've used 85% of your session limit” (errors).

One wrinkle: the Pro and Max Help Center page points to the /status command for monitoring, while the Claude Code documentation points to /usage. Try both on your version.

What drains the weekly limit fastest?

The costs documentation lists why a long session uses more of your plan than your activity suggests (why usage climbs). The ones that matter most for a coding agent:

• Long context re-read every turn. Claude Code sends the whole conversation with every request, and each tool call sends another. A one-line question late in a day-long session still draws usage for the whole history.
• Cold cache. The first message after a break longer than the cache lifetime reprocesses the full context.
• Subagents. Each subagent sends its own requests on top of the main conversation, and a subagent that inherits an expensive model spends at that model's rate. A switch to Opus also applies to subagents that inherit the session model.
• Verbose tool output. Test runs, build logs and file dumps land in context and are then re-read on every later request.
• Idle triggers. Scheduled tasks and cross-session messages fire while the session sits idle and send the full context each time.

On a paid plan, /usage recommends a change for any behaviour behind 10% or more of your recent usage.

How do I make the weekly limit last longer?

Start from the biggest slice in /usage, change one thing, and look again. The levers, in the order the cost documentation gives them (reduce token usage):

• /clear between unrelated tasks, and compact at natural pauses.
• Match the model to the job and lower effort on simple tasks.
• Keep verbose output out of the main context: delegate it or filter it. Our guide on long test output covers the filter route.
• Choose a smaller model for subagents that do simple work. See which model each subagent runs on.
• Avoid mid-task model switches: each model has its own prompt cache.

For the full diagnosis, read Claude Code usage limit running out fast.

What has Downshift measured, and what not?

Downshift is a hook that routes each subagent to a right-sized model and trims verbose command output. Its numbers are tokens per task and bytes, not subscription windows. Nobody has measured how many hours or days they add to a weekly limit, including us.

• Routing: four-arm battery, 2026-10-21, mechanical delegated tasks, 46 runs, one operator: about 24% fewer tokens per accepted task (16% per run; 95% interval +2.1% to +46.7%). Evidence page.
• Compaction: synthetic fixtures, 2026-10-21: 82.3% fewer bytes on the 26 of 248 files that compact, 23.7% across all 248. Bytes, not LLM tokens.

The interval is wide and the sample is small. Treat it as a direction, not a promise about your limit.

Related guides

• Claude Code usage limit running out fast: where the usage goes
• Claude Code subagent models: how the model is chosen
• Reduce Claude Code test output tokens, keep the failures
• Limite semanal do Claude Code explicadoPortuguês
• Install Downshift: Claude Code, Codex and Cursor hook setup

Sources

1. Anthropic Help Centerofficial
   support.claude.com: How do usage and length limits work
   session and weekly limits, shared use, no numbers (fetched 2026-10-24)
2. Anthropic Help Centerofficial
   support.claude.com: Using Claude Code with your Pro or Max plan
   Pro and Max limits, no figures, /status (fetched 2026-10-24)
3. Anthropic Help Centerofficial
   support.claude.com: Usage limit best practices
   Settings > Usage, weekly reset shown on the Usage page (fetched 2026-10-24)
4. Anthropic docsofficial
   code.claude.com/docs/en/costs
   /usage breakdown, why usage climbs, reduce token usage (fetched 2026-10-24)
5. Anthropic docsofficial
   code.claude.com/docs/en/errors
   limit messages, shared across models, session and weekly at once (fetched 2026-10-24)
6. Downshift repomeasured
   docs/evidence/four-arm-mechanical-2026-10-21.md
   46 runs, one operator; tokens per task, not subscription windows