TL;DR
- Anthropic introduces weekly rate limits starting August 28 for Claude Code users.
- Limits target Pro and Max plan subscribers, affecting less than 5% of users.
- The policy aims to curb 24/7 background usage and account reselling.
- Max users can buy more usage at standard API rates.
Usage Clampdown: A New Strategy for Claude Code
Anthropic is tightening controls on its AI-powered coding tool, Claude Code, by launching weekly rate limits. The policy, announced via email and Anthropic’s X account, will go into effect on August 28, targeting power users running the tool continuously and those sharing or reselling access — both violations of the company’s terms.
The limits apply to subscribers on the $20/month Pro, $100/month Max, and $200/month Max plans. These new caps are in addition to five-hour usage resets, which will remain in place.
Two New Limits: One General, One Opus-Specific
The company will implement two distinct seven-day usage limits:
- A general usage cap across models
- A model-specific cap for the high-performance Claude Opus 4
This means subscribers will face more predictable weekly constraints, designed to balance system availability for all users. Anthropic emphasized that most users will not be impacted, with only the top 5% of heavy users being affected.
“Claude Code has experienced unprecedented demand since launch,”
— Amie Rotherham, Anthropic Spokesperson
Rotherham added that these limits are necessary to maintain service reliability and are in response to recent outages, with the status page confirming seven disruptions in the past month.
Rate Tiers: How Much Access Do You Really Get?
According to Anthropic’s communication, here’s what users can expect per plan tier:
| Plan | Sonnet 4 Hours/Week | Opus 4 Hours/Week |
| Pro ($20/month) | 40–80 hrs | N/A |
| Max ($100/month) | 140–280 hrs | 15–35 hrs |
| Max ($200/month) | 240–480 hrs | 24–40 hrs |
While the company previously stated that the $200 Max plan provides 20x usage over the Pro tier, these new numbers reflect only about 6x more Claude Code hours, raising questions about how usage is measured. It’s unclear whether tokens, compute power, or another internal metric is the real baseline.
Context: Why AI Firms Are Rethinking Access
Anthropic’s move mirrors decisions by other AI tool providers grappling with resource limitations and unexpected demand surges. In June, Anysphere modified its pricing for Cursor, placing constraints on its $20/month Pro plan after abuse by power users. Following backlash, the company issued an apology for poor communication.
Likewise, Replit also made similar pricing changes for its AI coding environment earlier this year.
The broader issue: companies like Anthropic are constrained by GPU supply and are racing to deploy new AI data centers to meet growing model-serving demands. Most firms now walk a tightrope between scalability and accessibility.
Looking Ahead: Promises and Possibilities
Anthropic says it plans to support long-running coding tasks “through other options in the future.” Until then, users needing extra capacity must pay for overages at standard API rates.This move may allow Anthropic to focus on long-term platform stability, but it also highlights the underlying tension between product popularity and infrastructure readiness — a growing pain shared across the AI industry.
