Claude’s 5-Hour Usage Limit: The Warmup Timer Trick That Actually Works (2026)

1 How the 5-Hour Rolling Window Actually Works

Most vibe coders assume there is a daily reset, perhaps at midnight in their timezone, that clears their quota each morning.

There is not.

According to the Claude Help Centre, the limit is a rolling window, meaning messages you sent 5 or more hours ago gradually age out and stop counting against your quota.

The practical consequence is that two identical users, one who starts working at 7:00 AM and one who starts at 10:00 AM, will hit their limit and get a fresh window at completely different times.

This variability is the mechanism the warmup trick exploits.

Understanding which AI coding tool works best for your workflow depends partly on how these limits interact with your typical working hours.

2 The Clock-Hour Anchoring Mechanic Most People Miss

This is the detail most guides miss entirely, and it matters for precise scheduling.

According to both the claude-warmup repository and apidog.com’s analysis, Claude Code’s budget windows are floored to clock hours rather than calculated from the exact minute of the first message.

The practical implication: if you want a fresh window at 9:00 AM, schedule your warmup for 3:15 AM or 4:15 AM, not 3:00 AM or 4:00 AM.

Sending at the :15 mark ensures the warmup anchors to its intended clock hour with a small buffer.

For agentic workflows that run Claude Code for extended periods, this precision means the difference between a fresh window arriving mid-task or after a frustrating wait.

3 The Warmup Trick Explained

Nathan Onn, who documented the session overlap technique, describes it as “scheduling intelligently” within existing constraints rather than bypassing any rules.

No extra messages are gained, and your total quota stays exactly the same.

What changes is the timing of when your fresh window arrives.

The key is calculating how long before your real work starts you need to send the warmup message.

The rule of thumb is 2–4 hours before your typical start time, adjusted so the warmup falls outside the weekday peak window.

Vibe coders who ship consistently tend to have predictable daily routines, which makes this technique especially effective for them.

4 How to Automate the Warmup with claude-warmup

The repository, published by developer vdsmon, was forked 123 times and starred 125 times within weeks of publication, signalling how urgently the developer community needed this fix.

The warmup ping itself costs negligible tokens as a single “hi” message to Claude Haiku with no context uses an insignificant fraction of any session’s budget.

claude -p "hi" --model haiku --no-session-persistence

Setup Steps

1. 1
   Fork the repository Run gh repo fork vdsmon/claude-warmup --clone to create your own copy.
2. 2
   Generate an OAuth token Run claude setup-token on a machine already logged into Claude Code. The token begins with sk-ant-oat01-.
3. 3
   Store as a GitHub Secret Run gh secret set CLAUDE_OAUTH_TOKEN and paste your token. GitHub Actions will use this to authenticate.
4. 4
   Set your cron schedule Edit .github/workflows/warmup.yml. The default is 15 9 * * 1-5 (weekdays at 09:15 UTC). See the calculator below to find your ideal time.
5. 5
   Verify it worked Run /usage in Claude Code after the first scheduled run to confirm the window reset time has shifted to your intended hour.

For developers running Claude Code on a remote server or VPS, the GitHub Actions approach is preferable to a local cron job since GitHub’s runners operate independently of your machine being switched on.

According to the Claude Code official documentation, each scheduled task starts a full independent session and counts against the same usage limit as an interactive session, so the warmup does draw a trivial amount from your quota.

Source: Portkey.ai, Usagebar.com, 2026. Values are midpoints of published ranges (Pro: 40–80h, Max 5x: 140–280h, Max 20x: 240–480h).

If you prefer not to use GitHub Actions, Claude Code’s built-in agent mode supports scheduled routines that can trigger the same warmup ping directly from the desktop app without any external tooling.

5 When to Schedule: Peak Hours vs Off-Peak

In March 2026, Anthropic tightened limits during weekday peak hours in response to a surge in new users.

According to J.D. Hodges’s analysis, Claude reached the number one position on the US App Store for the first time that month, driven by a 30% month-on-month increase in web traffic, and Anthropic adjusted peak-hour capacity in response.

The off-peak promotion that had offered 2x usage during evenings and weekends ended on 28 March 2026, but off-peak hours remain meaningfully less congested than the weekday morning window.

The ideal warmup schedule for most UK and European vibe coders is a cron set to fire between 4am and 6am UTC on weekdays, which anchors the 5-hour window to a 4:00 or 5:00 AM start time and ensures a fresh window is available by the time the typical 9:00–10:00 AM working session begins.

For context on how large language models like Claude handle usage at scale, the peak-hour mechanics reflect real infrastructure constraints rather than arbitrary throttling.

6 What Happens to Your Weekly Cap

This is the most important caveat, and many guides that discuss the warmup trick omit it entirely.

Claude Code maintains two independent limit systems that operate simultaneously.

The weekly cap was introduced on 28 August 2025 and affects fewer than 5% of subscribers according to Portkey.ai’s reporting.

Heavy users hitting the weekly cap need either a plan upgrade or a reduction in token consumption per session.

Per Nicholas Rhodes’s analysis, settings.json configuration tweaks can reduce Claude Code’s token consumption by 60–80%, which significantly extends how long before the weekly cap is reached.

The context window also plays a role: Max plan subscribers have access to a 1M token context window, while Pro plan users remain capped at 200K tokens per session, meaning Pro users burn through weekly hours faster on equivalent tasks.

For reference on understanding the range of tools available, the vibe coding glossary covers both context windows and session limits in plain language.

7 Anthropic Does This Internally: The Warmup Is Official

This detail, documented in GitHub Issue #10392 on the official claude-code repository, fundamentally validates the technique.

The community’s warmup trick and Anthropic’s internal implementation are functionally identical: send a cheap, stateless Haiku message to anchor a session window before the expensive main model takes over.

The bug documented in Issue #10392 was not that Anthropic used a warmup message, but that it set parentUuid to null, which severed the connection to previous conversation history and broke the --continue flag.

The warmup mechanic itself was working as intended.

For vibe coders evaluating whether to use this technique, the fact that Anthropic’s own tooling relies on the same pattern is a meaningful signal that the approach is stable and supported rather than a loophole that may break in future releases.

Developers considering alternatives such as Cursor or other AI coding tools face different limit structures, but the session-anchoring principle is unique to Claude’s rolling window model.