Anthropic Briefly Suspended OpenClaw Creator's Account, Then Reversed Course
Peter Steinberger, the developer behind the popular open-source coding agent framework OpenClaw, woke up Friday to find his Anthropic account suspended. The company cited 'suspicious' activity. Steinberger, who now works at rival OpenAI, posted a screenshot of the ban notice on X with a warning: keeping OpenClaw compatible with Claude models just got harder.
The suspension was short-lived. Within hours, as the post racked up hundreds of replies and widespread attention, Steinberger reported his access had been restored. An Anthropic engineer even showed up in the replies, assuring Steinberger that the company has never banned anyone specifically for using OpenClaw and offering direct assistance.
The incident lands amid already strained relations between Anthropic and the open-source agent community. Just last week, Anthropic announced that Claude subscriptions would no longer cover third-party agent harnesses like OpenClaw, requiring developers to pay separately through the API based on usage. The company attributed the change to the heavy compute demands that coding agents place on its infrastructure — continuous reasoning loops, automatic retries, and extensive tool integrations all burn through resources far faster than ordinary chat prompts.
Steinberger was skeptical of that rationale. He noted the timing: Anthropic rolled out new features for its own Cowork agent, including Claude Dispatch for remote agent control and task assignment, just weeks before locking third-party tools out of subscription pricing. 'First they copy some popular features into their closed harness, then they lock out open source,' he wrote after the pricing change. Critics have taken to calling the new API-based pricing a 'claw tax.'
Steinberger's dual role made the episode even more charged. Some commenters suggested the ban was retaliation for his employment at OpenAI, a theory he didn't discourage. When one person told him he'd picked the wrong company, Steinberger fired back: 'One welcomed me, one sent legal threats.'
Still, he pushed back on the idea that he should simply stop using Claude. His work on OpenClaw, he explained, is separate from his day job at OpenAI. As head of the OpenClaw Foundation, his responsibility is making sure the framework runs well across every major model provider — and that means testing against Claude whether Anthropic likes it or not. The question now is whether Anthropic's evolving policies will keep making that job more difficult.
