Anthropic suspends OpenClaw creator’s access to Claude

"Yeah, folks, it's gonna be harder in the future to ensure OpenClaw still works with Anthropic models," OpenClaw creator Peter Steinberger wrote in a post on X early Friday morning, sharing a screenshot of a notice from Anthropic that said his account had been suspended due to "suspicious" activity.

The suspension turned out to be short-lived. Within a few hours, and after the post gained widespread attention, Steinberger said his account access had been restored. Among the many responses to his post — including speculation tied to his current role at OpenAI — an Anthropic engineer publicly replied, stating that the company does not ban users for using OpenClaw and offering assistance.

It remains unclear whether that interaction directly led to the account's reinstatement, but the exchange provided insight into recent tensions surrounding third-party tools and AI platforms.

The incident comes shortly after Anthropic confirmed that its Claude subscription plans would no longer cover usage through "third-party harnesses, including OpenClaw." Instead, users who rely on such integrations must now pay separately based on consumption via Claude's API. In effect, Anthropic — which has introduced its own agent system, Cowork — is now monetising external tool usage differently. Steinberger said he had already adjusted to the new system and was using the API as required, yet still encountered the suspension.

Anthropic has explained that the pricing change reflects how its subscription plans were not designed to support the heavier workloads associated with tools like OpenClaw. These systems often require more computing resources because they can run continuous reasoning loops, automatically repeat tasks, and integrate with multiple third-party services.

Steinberger, however, expressed scepticism about that reasoning. After the pricing update, he posted criticism suggesting that the timing was questionable, writing that it was "funny how timings match up," referencing what he saw as a pattern of introducing similar features internally and then limiting external alternatives. While he did not name specific features, the comment may refer to updates within Claude's Cowork system, including capabilities such as remote task execution tools introduced shortly before the pricing changes.

His frustration remained evident in his follow-up remarks on Friday. In one exchange, a user suggested that his current situation might be linked to his decision to work at OpenAI instead of Anthropic. Steinberger responded, "One welcomed me, one sent legal threats."

When questioned about why he continues to use Claude models despite working at OpenAI, Steinberger clarified that his usage is strictly for testing. He explained that his involvement with the OpenClaw Foundation focuses on ensuring compatibility across DI providers. At the same time, his professional role at OpenAI involves contributing to long-term product strategy.

Other users noted that Claude remains a preferred model among many OpenClaw users, which explains the need for continued testing. Steinberger acknowledged that feedback as well, responding briefly: "Working on that."