On April 4, 2026, at noon Pacific time, Anthropic flipped a switch that sent shockwaves through the open-source AI community. Claude Pro and Max subscribers can no longer use their flat-rate plans with third-party agent frameworks like OpenClaw. If they want to keep running autonomous agents on Claude, they have to pay separately through a new pay-as-you-go billing tier.

The math tells the whole story. A single OpenClaw instance running autonomously for a full day, browsing the web, managing calendars, responding to messages, and executing code, can burn through $1,000 to $5,000 in API-equivalent costs. Under a $200-per-month Max subscription, that was an astonishing transfer of compute costs from users to Anthropic. It was never going to last.

But that does not mean the timing was not brutal.

135,000 Instances and a Very Big Bill

At the time of the announcement, an estimated 135,000 OpenClaw instances were running on Claude subscriptions. These were not hobbyists playing around. Many were developers who had built entire personal AI infrastructures on the assumption that their monthly subscription covered it. Some had been running agents 24/7 for months. The overnight switch to pay-as-you-go pricing means some of these users face cost increases of 50x or more.

Boris Cherny, Head of Claude Code at Anthropic, framed it as a capacity management decision. "Anthropics subscriptions were not built for the usage patterns of these third-party tools," he told TNW. "Capacity is a resource we manage thoughtfully and we are prioritising our customers using our products and API."

Translation: the price gap between what heavy agent users paid under flat subscriptions and what equivalent usage would cost at API rates was more than 5x. Anthropic was quietly subsidizing a use case it never priced for, and it could not afford to keep doing it.

The Open Source Community Is Furious

Peter Steinberger, the Austrian developer who created OpenClaw before selling it to OpenAI in February 2026, called the decision a betrayal of open-source developers. The Hacker News thread about the change is exactly as angry as you would expect. Reddit is worse.

The frustration is understandable. Anthropic built its brand on being the developer-friendly alternative to OpenAI. Claude Code was specifically marketed to the kind of power users who build agent workflows. Those users are now being told that the thing they built their workflows around was never actually meant for them.

What Happens Next

Anthropic says it will extend the restriction to all third-party harnesses in the coming weeks. This is not a one-time correction. It is a permanent policy shift. The era of running autonomous AI agents on flat-rate subscriptions is over, at least at Anthropic.

The irony is that OpenAI, which now owns OpenClaw, has every incentive to make its own subscriptions work seamlessly with the platform. Anthropic just handed its biggest competitor a loyalty gift wrapped in developer rage.

Reporting based on TNW, VentureBeat, TechCrunch, and Hacker News community discussion.