«Yes, folks, it will be more difficult to get OpenClaw to work with Anthropic models in the future,» OpenClaw creator Peter Steinberger wrote on X early Friday morning, attaching a photo of a message from Anthropic saying his account had been suspended due to «suspicious» activity.
«Yes, folks, it will be more difficult to get OpenClaw to work with Anthropic models in the future,» OpenClaw creator Peter Steinberger wrote on X early Friday morning, attaching a photo of a message from Anthropic saying his account had been suspended due to «suspicious» activity.
The ban didn’t last long. A few hours after the post went viral, Steinberger announced that his account had been reinstated, TechCrunch reports .
Among the hundreds of comments — many of which bordered on conspiracy theories, given that Steinberger now works at OpenAI, an Anthropic competitor — was a comment from an Anthropic engineer, who assured the prominent developer that the company had never banned users for using OpenClaw, and offered to help.
Yeah folks, it’s gonna be harder in the future to ensure OpenClaw still works with Anthropic models. pic.twitter.com/U6F8GZvPcH
— Peter Steinberger 🦞 (@steipete) April 10, 2026
This lockout follows recent news that Anthropic’s Claude subscriptions will no longer cover the use of «third-party wrappers, including OpenClaw.»
OpenClaw users now have to pay for such usage separately via the Claude API, based on actual consumption. In essence, Anthropic, which offers its own agent called Cowork, is now charging a «Claw tax.» Steinberger noted that he followed this new rule and used his own API, but was still blocked.
Anthropic said it made the pricing changes because subscriptions were not designed for the «use cases» of such shells. Such tools can be more resource-intensive than regular queries or simple scripts, as they can run continuous reasoning loops, automatically repeat tasks, and integrate with a wide range of other third-party services.
Steinberger, however, did not buy that excuse. After Anthropic changed its pricing, he wrote, «It’s interesting how the timing matches up: first they copy popular features into their own closed-source wrapper, then they lock down the open source code.» Although he did not elaborate, he likely meant features added to the Claude Cowork agent—such as Claude Dispatch, which allows users to remotely manage agents and assign tasks. Dispatch was launched a few weeks before Anthropic changed its pricing policy for OpenClaw.
Steinberger’s frustration with Anthropic spilled out again on Friday.
One user implied that the developer was to blame for this situation, as he chose to work at OpenAI rather than Anthropic, writing: «You had a choice, but you went to the wrong ones.» To which Steinberger replied: «Some welcomed me, while others sent legal threats.»
A hurtful remark.
When many people asked him why he was using Claude at all instead of his employer’s models, he explained that he was doing it solely for testing purposes—to make sure that OpenClaw updates wouldn’t cause problems for Claude users.
He explained, «There are two things to keep in mind. My work at the OpenClaw Foundation, where we strive to make OpenClaw work great with any model vendor, and my position at OpenAI, where I help them with their future product strategy.»
Many also noted that the need to test Claude arises because the model remains a popular choice among OpenClaw users, even compared to ChatGPT. He was also reminded of this when Anthropic changed its pricing policy, to which he replied: «I’m working on it.» (And this is a hint at what exactly his responsibilities at OpenAI are).
