Anthropic updated its pricing page on April 21 to remove Claude Code from the feature list of its $20/month Pro subscription plan, replacing the checkmark with an explicit “X.” The change, first flagged by AI industry commentator Ed Zitron, triggered immediate backlash from developers before Anthropic clarified it was a limited experiment, according to The Register.
What Actually Changed
On Monday, the claude.com/pricing page stated the Pro plan “includes Claude Code.” By Tuesday, that phrase was gone. But the rollout was inconsistent: at the time of The Register’s reporting, the Claude Code product page still listed Pro plan access, and the CLI terminal output still displayed “Claude Pro” when accessed, per The Register.
Anthropic’s head of growth Amol Avasare posted on X to clarify: “For clarity, we’re running a small test on ~2 percent of new prosumer signups. Existing Pro and Max subscribers aren’t affected.”
Why Anthropic believed a public pricing page update would only reach 2% of users and not the broader internet was not explained. A company spokesperson did not respond to The Register’s follow-up question on that point.
The Capacity Problem
Avasare’s full explanation laid out the business logic directly: “When we launched Max a year ago, it didn’t include Claude Code, Cowork didn’t exist, and agents that run for hours weren’t a thing. Max was designed for heavy chat usage, that’s it.”
Since then, Claude Code was bundled into Max and “took off after Opus 4.” Cowork launched. Long-running async agents became daily workflows. Per Avasare: “Engagement per subscriber is way up. We’ve made small adjustments along the way (weekly caps, tighter limits at peak), but usage has changed a lot and our current plans weren’t built for this.”
The core issue, as The Register reports, is that Anthropic’s subscription plans charge far less than the book value of tokens consumed, sometimes by a factor of ten or more. Projects like OpenClaw led many developers to create long-running agents using Claude models, compounding the cost gap between subscription revenue and compute spend.
Pricing Agents as Premium Products
The test, limited as it may be, signals Anthropic’s direction. Claude Code is the company’s most differentiated product for software engineering, handling multi-file edits, testing, deployment, and complex refactoring autonomously. Gating it behind the $200/month Max tier would position agentic coding as a premium capability with pricing that reflects actual compute costs.
This mirrors a broader pattern. The New Stack and The Register both note that GitHub and Google have faced similar capacity constraints from agentic workloads. GitHub recently grounded some Copilot accounts, and Google’s Gemini CLI has faced usage discussions in its developer community.
The 2% test may be small, but the question it is probing is large: can subscription-based AI products survive when agentic workloads consume 10x the compute of chat-based usage at the same price point? Anthropic’s answer appears to be restructuring, not just rate limiting.