Anthropic · 2026-04-21 · major
Anthropic Quietly Drops Claude Code From New Pro Signups in A/B Test
Anthropic stripped Claude Code from the $20/mo Pro plan on its pricing page on April 21. Head of Growth Amol Avasare clarified it's a test on ~2% of new prosumer signups; existing Pro and Max subscribers are unaffected. The change is the second flat-rate AI plan restructuring this week after GitHub Copilot.
Anthropic is testing whether $20/mo Pro can survive without Claude Code — the most-used reason to subscribe.
Key specs
| Test scope | ~2% of new prosumer signups |
|---|---|
| Pro plan price | $20/month |
| Existing subscribers affected | 0 |
What is it?
On April 21, Anthropic's pricing page quietly stopped listing Claude Code as part of the Pro plan: Monday it said 'Includes Claude Code', Tuesday it showed an explicit X next to that feature for the Pro tier. Following developer backlash on X and Reddit, Anthropic's Head of Growth Amol Avasare confirmed the change is an experiment running on roughly 2% of new prosumer signups — current Pro and Max subscribers are not affected, and Claude Code remains in Max.
How does it work?
When Max launched a year ago it didn't include Claude Code; the feature was bundled later and exploded after Opus 4. Per Avasare, engagement per Pro subscriber has shifted dramatically as users run long-horizon agentic coding sessions, and the per-session compute on those sessions blows past flat-rate plan economics. Rather than raise the price, Anthropic is A/B testing what the new-signup funnel looks like with Claude Code unbundled and pushed up to the $100/mo Max tier.
Why does it matter?
This is the second flat-rate AI plan restructuring in a single week — GitHub Copilot just paused new Pro signups and pulled Opus from the Pro tier on April 20. Together they're a strong signal that flat-rate $20 plans cannot economically support agentic workloads at the current pricing, and that the entire AI subscription model is repricing in real time. For developers on Pro, it's also the strongest argument yet for local models (Qwen3.6, Kimi K2.6, GPT-OSS) and self-hosted coding agents (OpenCode, Cline, Aider).
Who is it for?
Anyone on a Claude Pro subscription or considering one for Claude Code; anyone tracking the economics of agentic AI subscriptions.
Try it
Check claude.com/pricing/pro — if you're in the test group, Claude Code shows an X under Pro features.