Anthropic has quietly removed its agentic Claude Code feature from the $20 Pro plan, repricing it at $100 or more per month , and developers are not taking it well.
As of April 21, 2026, if you want to use Claude Code , Anthropic’s autonomous coding tool that can write, debug, and execute software independently through a terminal interface , you’ll need to open your wallet considerably wider than before. What was bundled into a $20 monthly subscription is now gated behind a usage-based tier starting at $100, a fivefold jump at minimum and a tenfold one for anyone who was getting meaningful mileage out of the feature. The announcement landed fast, with changes appearing in user dashboards almost immediately. The reaction on Reddit and X has been equally swift and considerably less polished.
Developers are using the word ‘bait-and-switch’ with striking consistency across community threads. The pattern they’re describing is familiar in tech: a compelling feature rolls out at an accessible price point, builds a loyal base, then gets repriced toward enterprise. Whether or not that framing is entirely fair to Anthropic, perception in developer communities tends to harden quickly, and the word ‘switcher’ is already circulating as a self-identifier among users announcing their return to OpenAI.
Strip away the user frustration and there’s a clear strategic statement underneath this move. Anthropic is drawing a firm line between consumer AI and agentic AI. In the company’s emerging worldview, autonomous agents that take real actions , running code, navigating systems, completing multi-step tasks without hand-holding , belong in an enterprise-grade category, priced accordingly. That’s a coherent thesis. Agentic capability requires more compute, more guardrails, and carries more liability than a chatbot generating a cover letter. The $100 floor reflects that math, at least internally.
The problem is that Anthropic built its developer credibility on the back of Claude 3.7 Sonnet, a model that earned genuine respect in coding benchmarks and attracted technically sophisticated early adopters who are precisely the crowd least likely to absorb a surprise price increase without investigating alternatives. These are not passive consumers. They benchmark, they compare, they publish results, and they migrate with visibility.
OpenAI’s accidental windfall
OpenAI has not announced any formal response to this situation, and it doesn’t need to. The discontent is doing the recruitment for them. GPT-4 remains available at a significantly lower price point and continues to deliver capable code generation, even if it lacks a purpose-built agentic terminal equivalent to what Claude Code offered. For a developer who was paying $20 for Claude Pro and used Claude Code heavily, the calculus for switching is almost frictionless right now. The only question is whether OpenAI moves to capitalize more explicitly , perhaps by accelerating or announcing an agentic coding feature of their own at a competitive price.
That’s the scenario Anthropic should be most concerned about. Developer mindshare is hard to win and easy to lose. Anthropic gained significant ground with Claude 3.7 Sonnet’s coding performance, creating genuine enthusiasm in communities that had historically been OpenAI-first. That goodwill is now under stress in a way that a blog post explaining the pricing rationale is unlikely to fully repair.
The broader fracture in the AI market
What this moment reveals is a fracturing of the AI developer market along willingness-to-pay lines that nobody had fully mapped yet. There’s a consumer tier , casual users, writers, light coders , for whom $20 remains a reasonable ceiling. Then there’s a professional developer segment that demands agentic capability, cares deeply about pricing, and has real alternatives. Anthropic has now explicitly priced those two groups differently. The risk is that the professional segment, rather than accepting the new price, simply leaves.
The companies that win developer loyalty in the next 12 months will likely be those that find a middle tier , something above $20 but well below $100 , that covers serious coding use without requiring an enterprise procurement conversation. Watch whether OpenAI or a smaller player like Mistral or Cursor moves into that gap. If someone credible does, Anthropic’s repricing decision will look considerably more costly in retrospect than it does today.
Also read: OpenAI’s GPT Image 2 rewrites the rules of visual AI and puts creative software giants on notice • Anthropic tests stripping Claude Code from its base plan and OpenAI employees are already laughing • Mozilla used Anthropic’s Mythos to find and fix 271 bugs in Firefox, signaling a new era for AI-driven software quality