Editor's Note: Recently, a change on Anthropic's pricing page has sparked widespread attention in the developer community: the Claude Code, originally included in the Pro subscription, suddenly disappeared, with some pages even marking this feature as no longer included with an "X." Although the company later responded that this was only a small-scale test for about 2% of new users and existing subscribers were not affected, inconsistencies in information across different pages, coupled with misalignments between customer service and the actual product status, quickly magnified this adjustment into a signal of "benefits tightening."
From the event itself, it appears more like a trust fluctuation caused by communication disorder; but in the broader industry context, it reflects a structural shift that AI products are undergoing. As explained by Amol Avasare, users' use of Claude has evolved from simple conversations to code generation, collaborative development, and even long-running agent workflows, significantly increasing per-user resource consumption. However, the existing subscription system is still built on the assumption of "light usage," struggling to cover the rapidly expanding computing costs.
This also explains why Anthropic has recently introduced usage restrictions frequently and explored new package structures — as model capabilities continue to improve and the developer ecosystem keeps expanding (such as projects like OpenClaw), the platform has had to find a new balance between "affordability" and "sustainability." This pressure is not unique, as Google, GitHub, and others are also facing similar issues.
When AI transitions from a tool to infrastructure used frequently and heavily, does the existing subscription logic still hold? And how will platforms redefine "payment" between cost constraints and user experience.
The following is the original text:
Anthropic has removed Claude Code from its Pro subscription plan — at least as far as some public-facing web pages indicate. However, the company stated that this was only a test for a small number of users.
On Monday, the company's pricing page stated that the Pro plan "includes Claude Code"; but by Tuesday, this statement had disappeared, and on the same page's feature list, the Claude Code item changed from a checkmark to a clear "X." This change was first pointed out by AI industry skeptic Ed Zitron.
However, this product adjustment has not been synchronized across all of Anthropic's pages. As of the time of writing this article, Claude Code's product page still shows that the Pro plan has access to the code generation tool. When this journalist accessed Claude Code via the command-line interface (CLI), the terminal output still referred to it as "Claude Pro."
When asked about this, Claude.ai insisted that the Pro plan includes Claude Code. However, in a document page referenced today, only the Max plan is mentioned.
Losing access to Claude Code has evidently left many developers uneasy, and this sentiment has reached Anthropic. Amol Avasare, the company's Head of Growth, posted on social media, stating that this adjustment was actually an experiment.
He said: "To clarify, we are running a small-scale test on approximately 2% of newly registered Pro (for professional users) users. Existing Pro and Max subscription users will not be affected."
As for why Anthropic believed that updating its public pricing page would only impact some users rather than the whole internet, it is currently unclear. The company's spokesperson has also not responded to follow-up inquiries on this matter.
Amol Avasare further explained: "When we launched Max about a year ago, it did not include Claude Code, Cowork did not exist, and agents capable of running for hours on end had not appeared. The design goal of Max at that time was simple, to cater to high-frequency chat usage. However, since then, we bundled Claude Code into Max, and it quickly gained popularity after Opus 4 was released. Cowork went live, and long-running asynchronous agents became a part of daily workflows. The actual usage of Claude subscriptions by users has fundamentally changed. The intensity of usage by each subscription user has significantly increased. We have made some minor adjustments along the way (such as weekly quotas, tightening limits during peak times), but the overall usage patterns are very different now, and the existing packages were not designed for this scenario. So, we are exploring different solutions to continue providing a good experience for users. The specific form is not yet determined — this is exactly what we are currently testing and collecting feedback on."
Recently, Anthropic has been taking measures to limit the usage of its AI models to mitigate demand during a period of tight compute availability, a challenge also faced by GitHub and Google. Over the past few months, the Claude series of models has gained rapid popularity, and with projects like OpenClaw emerging, many developers have started building various agents based on Claude.
At the core of the issue is that Anthropic's subscription fees are far below the actual token cost of consumption, sometimes even by an order of magnitude. Last month, the company introduced usage restrictions, encouraging users to utilize the service during off-peak hours, similar to how utility companies incentivize energy conservation through time-of-use pricing. As Avasare mentioned, Anthropic is exploring further usage limitation schemes.
Unsurprisingly, many Claude users expressed their dissatisfaction on the forum. Several believed this was a substantive pricing plan adjustment and pointed out that Anthropic seemed to be "covertly" steering original Pro users towards lower-cost Chinese models such as Minimax, Qwen, Kimi, and GLM.
More notably, it is not the change itself that is causing concern, but rather the lack of communication. Anthropic seemingly did not clearly communicate the adjustments to users, behaving as though minimal notification was sufficient for individual users paying only $20 per month.
However, enterprise customers are evidently less tolerant of this ambiguity, potentially raising concerns that Anthropic does not prioritize customer relationships.
Avasare appears to have realized this as well. He stated, "Once we finalize the solution, if there are changes for existing subscription users, everyone will be notified well in advance." He pledged that users "will hear directly from us, not through screenshots on X or Reddit."
It is noteworthy that this statement itself was made by him on X.
