in August 2025, Antropic introduced weekly usage limits to its cloud code, and in January 2026, it blocked circumvention access via third-party tools altogether, because some power users were consuming more than $1000 per month in API usage on a subscription, causing service outages. Antropic says it only affects less than 5% of all subscribers, but the developer community has criticized the lack of transparency and unannounced policy changes.

why did the unlimited subscription fall apart?

according to Antropic's official announcement on July 28, 2025, some power users had been monopolizing system capacity by running the cloud code continuously in the background 24/7. one user consumed a whopping $13200 worth of model usage on a $200 subscription and was also caught sharing and reselling accounts. a spokesperson for Antropic admitted that Claude Code had experienced unprecedented demand since its launch, with at least seven service outages in the past month.

from an economic perspective, the old model was also unsustainable: developers were consuming $120 worth of API credits per day on the $200 Max plan, which equates to $3600 per month, or 18 times the value of the full price.

details of the new acceptable use policy

the new policy, which went into effect on August 28, 2025, is a dual structure that adds a 7-day weekly limit on top of the existing 5-hour rolling limit. Opus 4 models are subject to a separate, stricter sub-limit.

weekly usage limits by plan

attribute Pro plan Max 5x plan Max 20x plan monthly Price 20 100 200 Sonnet 4 Weekly Limits 40 to 80 hours 140 to 280 hours 240 to 480 hours Opus 4 Weekly Limit not available 15 to 35 hours 24 to 40 hours if limit is exceeded wait until next week Additional API fees can be purchased Additional API charges can be purchased

it's worth noting that the limits are presented as ranges. antropic explained that the actual time available depends on message length and attachments and how the feature is used.

on January 9, 2026, a stronger measure was added. the Antropic technical team confirmed that they have strengthened protection measures against Claude Code Harness impersonation. They blocked third-party tools, including OpenCode and Cursor customizations, from accessing the API by posing as a Claude Code client.

claude Code's stagnation and explosive growth

claude Code is Antropic's terminal-based agentive coding tool. it debuted in a limited research preview with Claude 3.7 Sonnet in February 2025 and became generally available in May of the same year with the release of Claude 4.

claude Code core features

property content interface terminal and CLI-based key features read and edit files, execute terminal commands, create Git commits, refactor multi-files integration support Integrates Figma with Jira and Slack via the MCP protocol IDE integration Support for VS Code extensions and Xcode 26 integration specialized features background agent mode, learn project-specific context from CLAUDE.md files

growth is explosive. in November 2025, Antropic's Chief Product Officer announced that Claude Code had reached $1 billion in annualized revenue in just six months, representing 14 percent of Antropic's total estimated revenue of $7 billion.

developer community reacts with excitement

the Korean developer community reacted to the news by saying that 1 in 20 people go over the limit, not less than 5 percent, which by SaaS industry standards is the majority of real users. The biggest complaint is that there is no way to see how much usage is left. one user wrote that not being able to see how much you've spent creates a sense of rationing.

in the global developer community, the thread was hotly debated, with 609 points and 705 comments. the Ruby on Rails founder criticized the decision as very hostile to customers. Some users claimed that their weekly usage plummeted from 40 to 50 hours to 6 to 8 hours.

on the other hand, there are also voices of understanding. one expert countered that the backlash against Antropic was silly, saying that if you open a buffet restaurant and some people eat 10 times as much every day, it's bound to be a problem. Another user said that a few people running agents around the clock is better than the majority not being able to use the service.

competitor analysis

Claude Code's direct competitors in the AI coding assistant market are GitHub Copilot and Cursor.

AI coding tools comparison table

attribute claude Code GitHub Copilot Cursor basic pricing 20 per month Pro 10 per month Pro 20 per month Pro highest Rate 200 per month Max 20x 39 per month Pro+ 200 per month Ultra interfaces terminal and CLI IDE extensions standalone IDE multi-file editing strong support limited Excellent with Composer context windows up to 200K tokens 8192 tokens maximum supported model selection Claude only GPT with Claude and Gemini with o3 GPT with Claude and Gemini with Grok

GitHub Copilot offers the cheapest entry point at $10/month, but has limited contextual understanding. Cursor is strong in multi-file Composer editing, but faced similar controversy when it switched from request-based to credit-based pricing in June 2025. claude Code still has a technical edge in complex reasoning, architectural decisions, and code reviews.

industry-wide, the end of the era of unlimited AI is being heralded. Cursor, Replit, and others have made similar moves to usage-based billing, and managed consumption is expected to become the standard for AI services by 2026.

what developers should do

this move signals the maturation of the AI coding tools market. while Antropic's decision was economically inevitable, its lack of transparency and failure to communicate upfront undermined developer trust.

while hobbyists and developers with typical usage patterns won't be affected, power users running around-the-clock automation or parallel instances will need to switch to using the API directly or consider open source alternatives like Qwen3-Coder.

in the long term, building local LLM environments is likely to emerge as an alternative. one developer predicted that there will come a time when 99 percent of daily tasks will be handled locally and without restriction. antropic has stated that it will provide other options in the future to support long-running use cases, but no specific enterprise alternatives have been revealed.

frequently asked questions

Q1. When will the load code usage limits go into effect?

A1. Weekly usage limits went into effect on August 28, 2025. the blocking of third-party tools went into effect on January 9, 2026.

Q2. Are general users affected by the limits?

A2. According to Antropic, less than 5% of all subscribers are affected. Most users doing typical development tasks will stay within the limits.

Q3. What happens if I exceed the limit?

A3. Pro plans will have to wait until the following week. Max plans can purchase additional usage at standard API rates.

Q4. Can I check my usage?

A4. There is currently no official way to check usage, which is the biggest complaint from developers.

Q5. What are the alternatives to Claude Code?

A5. GitHub Copilot is great if you want a low-cost entry point. Cursor is strong for multi-file editing. if you need unlimited usage, there are also open source models like Qwen3-Coder that you can run locally.

conclusion and CTA

the days of unlimited Claude Code subscriptions are over, and developers need to examine their usage patterns and understand the limits of their plans to best utilize them.

if you found this article helpful, let us know what you think in the comments. to stay up to date on the latest AI development tools, subscribe to our blog.