Anthropic quietly A/B tested pulling Claude Code from Pro tier, then backtracked

Anthropic ran an unannounced pricing experiment on roughly 2% of new Pro-tier signups, gating Claude Code behind the $100/month Max plan. The test caught wider attention because the public pricing page was updated to show Claude Code as universally excluded from Pro, creating the impression of a permanent tier change rather than a limited trial. Existing Pro subscribers retained access; only new prosumer signups in the test bucket were blocked.

Head of growth Amol Avasare framed the move as a response to shifting usage economics. When Max launched a year ago, it was positioned for heavy chat workloads — Claude Code wasn’t bundled, Cowork didn’t exist, and multi-hour agent runs weren’t a common pattern. The bundling of Claude Code into Max after Opus 4, combined with long-running async agents becoming routine, has driven engagement per subscriber sharply upward, straining plans that weren’t designed for agentic workloads.

The rollback came quickly: the pricing page was reverted to show Claude Code included in Pro. The episode signals that Anthropic’s current tier structure is economically misaligned with how subscribers actually consume compute, and further repricing — weekly caps, peak-hour throttling, or outright tier restructuring — is likely on the roadmap.