Claude Code Rate Limits Explained (2026)

Claude Code rate limits are one of the most common sources of frustration for developers. You are in the middle of a productive session, and suddenly Claude tells you to come back later. Understanding how these limits work is the first step to managing them effectively.

This post explains the rate limit system as it works today, based on what Anthropic has disclosed and what users can observe through the API.

The Two-Window System

Claude Code uses two overlapping rate limit windows. Both must stay below their limits for you to continue using the service.

The 5-Hour Rolling Window

The primary rate limit is a 5-hour rolling window. This tracks your cumulative usage over the past 5 hours as a utilization percentage. When you hit 100%, you are rate limited.

Important details:

• It rolls continuously. It is not a fixed block from 9am to 2pm. If you start using Claude Code at 10:30am, usage from 10:30am starts dropping off the window at 3:30pm.
• Recovery is gradual. As older usage ages past the 5-hour mark, your utilization percentage decreases. You do not go from 100% to 0% at once - it decreases as each piece of usage drops off.
• Burst usage hits hard. A single intensive hour can push you past the limit quickly. The window does not reward spreading usage evenly within it - any usage within the past 5 hours counts equally.

The 7-Day Rolling Window

The secondary rate limit is a 7-day rolling window. This works the same way as the 5-hour window but over a much longer period. It prevents users from consuming too much capacity over the course of a week.

This limit is less commonly hit during normal usage, but it catches sustained heavy usage. If you are pushing Claude Code hard every day for a week, you may hit the 7-day limit even if you never hit the 5-hour limit in a single session.

Recovery from the 7-day limit is slower for the same reason - you need to wait for usage from days ago to age out of the window.

Utilization Percentages, Not Token Counts

One of the most common misconceptions is that Claude Code has a fixed token limit - for example, "50,000 tokens per 5 hours." This is not how it works.

Anthropic does not publish specific token counts for rate limits. Instead, the system reports utilization as a percentage. When you check your rate limit status (through the API or tools like My Claude), you see something like "5h: 45%, 7d: 22%."

The actual capacity behind those percentages can vary. Anthropic adjusts allocations based on system load, plan tier, and other factors. The percentage is the only reliable metric - it tells you exactly how close you are to the limit regardless of what the underlying numbers are.

This is why tools that try to count tokens locally can give misleading results. Your local token count might suggest you should have plenty of headroom, but the API's utilization percentage is the ground truth.

Plan Tiers

Pro Plan - $20/month

The Pro plan includes Claude Code access with standard rate limits. For developers who use Claude Code a few hours a day, Pro is often sufficient. However, intensive sessions or heavy daily usage can hit the limit, especially on the 5-hour window.

Max Plan - $100/month

The $100/month Max plan provides substantially higher rate limits. If you are regularly hitting Pro limits, this tier gives you significantly more room. It is designed for developers who use Claude Code as a primary tool throughout their workday.

Max Plan - $200/month

The $200/month Max tier offers the highest rate limits available. This is for power users who rely on Claude Code for hours every day and need the maximum possible allocation.

The jump between tiers is significant. If you are on Pro and frequently rate limited, upgrading to Max at the $100/month level typically resolves the issue for most usage patterns.

What Happens When You Hit the Limit

When either window reaches 100% utilization:

1. Claude Code tells you. You see a message indicating you have been rate limited, with a suggestion to try again later.
2. No new prompts. You cannot send new messages to Claude until utilization drops.
3. No extra charges. You are not billed more for hitting the limit. It is a usage gate, not a billing event.
4. Existing sessions are fine. Your conversation history is preserved. You can resume when the limit resets.

The 5-hour window typically recovers within 30 minutes to 2 hours, depending on how much usage was clustered. The 7-day window recovers as daily usage from a week ago drops off.

Monitoring Your Rate Limits

The built-in /status command gives you a snapshot, but for ongoing visibility, dedicated tools are more practical.

My Claude is a free, open-source desktop app that shows your utilization percentages in real time for both windows. It also provides:

• Threshold alerts before you hit the limit
• Pace predictions that estimate whether your current rate will cause a limit hit
• Historical tracking to identify usage patterns
• Desktop, WhatsApp, and Telegram notifications

Having this data visible at all times helps you make informed decisions about when to push hard and when to ease off.

Key Takeaways

• Claude Code has two rate limit windows: 5-hour and 7-day
• Limits are reported as utilization percentages, not fixed token counts
• The 5-hour window is the primary constraint for most users
• Pro, Max $100, and Max $200 plans have progressively higher limits
• Hitting the limit is temporary - recovery is automatic as usage ages out
• Monitoring tools help you avoid surprises

Understanding the rate limit system is the first step. The next step is managing your usage proactively so you rarely hit the limit in the first place.