OpenClaw founder Peter Steinberger (steipete) revealed on X on April 10 that Anthropic has blocked his API key, preventing him from running end-to-end tests on Anthropic models anymore. He warned: “In the future, ensuring OpenClaw can still work with Anthropic models will become even more difficult.”
Dual block of both the API and the CLI
When community members asked whether it was only blocked for the API, steipete confirmed that “both have been blocked.” He explained that the OpenClaw team continues to run end-to-end tests to ensure that each harness update won’t break compatibility with mainstream models. But now that they’ve lost Anthropic’s key, this process can no longer continue.
Tension between open-source competitors and the platform
Anthropic’s action of blocking OpenClaw developers’ accounts reflects the growing tension between AI platform providers and open-source competitors. As an open-source alternative to Claude Code, OpenClaw has long supported multiple model providers, including Anthropic’s Claude. But from Anthropic’s perspective, a competitor using its own API to ensure compatibility is, in essence, leveraging Anthropic’s resources to strengthen its competing offering.
This aligns with broader industry trends. As competition in the AI tools market intensifies, model providers are re-evaluating their API policies—how to strike a balance between an open ecosystem and protecting competitive advantages. For OpenClaw users, this could mean running into more compatibility issues when using Anthropic models in the future.
steipete previously said in a tweet that the latest version of OpenClaw has added support for inferrs (a new high-performance TurboQuant inference server), showing that the team is actively expanding support for local models—potentially a strategic response to hedge against the risk of platform blocking.
This article, “Anthropic blocks OpenClaw developer accounts; steipete: supporting Anthropic models will be more difficult in the future,” first appeared on ChainNews ABMedia.