Anthropic has abruptly halted a planned shift to token-based billing for its Claude Agent SDK, reversing course just as the changes were set to take effect.
The new pricing model had been scheduled to roll out on June 15. According to Ars Technica, the change would have heavily increased costs for power users of the SDK — the developer toolkit that lets engineers build automated AI agents powered by Claude.
Instead, Anthropic announced on Monday that it was pausing those pricing changes. The company offered a brief reassurance to developers: "Nothing changes for now," according to reporting by Ars Technica's Kyle Orland.
The SDK in question is designed for building multi-step, autonomous AI workflows — programs that can browse the web, write code, or complete tasks without constant human input. Billing for such tools is complicated because agent tasks can consume far more tokens than a simple back-and-forth chat, making pricing decisions especially consequential for businesses building on top of the platform.
Anthropic did not offer a detailed explanation for the pause or indicate when, or whether, a revised billing structure might be introduced.
The episode is a reminder that as AI companies rush to commercialize powerful developer tools, pricing decisions can face significant pushback — and that even well-signaled changes can get pulled at the last minute when the stakes for customers are high.