Anthropic quietly removes Claude Code from $20 Pro plan, developers revolt

Anthropic has quietly removed Claude Code from its $20 Pro plan pricing page without any announcement.

Users noticed that on the http://claude.com/pricing feature comparison table, the Claude Code column for the Pro plan now shows an X. Meanwhile, multiple help documents have deleted the phrase "Pro plan includes Claude Code"—though older versions can still be seen on the Wayback Machine.

But Anthropic's own Claude Code product page still says "Pro plan includes," and the customer service bot insists it's included. Having two contradictory statements on the same company website is a problem in itself. Some Pro users report Claude Code still works for them, but it's unclear whether this is a gradual rollout or if existing users are being grandfathered in.

If this change is real, the minimum monthly cost for developers to access Claude Code jumps from $20 to $100.

Viewed in a broader context, this move isn't surprising. Over the past two months, Anthropic has been tightening compute supply: first by restricting third-party tools (like OpenClaw) from using subscription credits, then by cutting quotas during peak hours, plus the fact that Opus 4.7 consumes significantly more tokens than its predecessor. Many Pro users report burning through their weekly allowance by Wednesday. All signs point to the same conclusion: Anthropic's inference compute is stretched thin, and they're prioritizing limited resources toward higher-paying users and enterprise clients.

Developers on Hacker News are furious. Many say they're planning to switch to OpenAI's Codex (still included in the $20 plan), or even use Chinese models like Kimi, GLM, and MiniMax, which offer lower prices and higher quotas. Others mention that Cursor's $20 plan with Composer 2 provides a decent experience. Another perspective is that Anthropic was probably losing money on the Pro plan anyway, so cutting it makes sense—but the execution is sloppy: quietly changing the page, no announcement, and a customer service bot still saying the opposite.

One user on the $200 plan did the math: he consumes about $3,000 worth of API costs each month. That explains why Anthropic can't sustain this, but it also shows that even high-paying users aren't loyal customers—"If OpenAI's model gets better, I'm gone immediately."

For Anthropic, the trickiest problem might not be short-term revenue, but the network effect of individual developers within companies. Many businesses adopted Claude because developers tried it first, liked it, and then pushed for corporate procurement. Alienating this group could damage the long-term enterprise sales pipeline.

Of course, some point out that OpenAI's current generosity with Codex on the $20 plan might not last forever—that level of spending probably isn't sustainable either. What Anthropic is doing now, OpenAI might do later. The real question is: who can keep their compute running longest will survive the price war.