Anthropic will block subscribers to its Claude Pro and Claude Max plans from integrating third-party AI agent tools, VentureBeat reported on Friday.

Boris Cherny (보리스 체르니), head of Anthropic's Claude Code, wrote on social media platform X (Twitter) that the company had been working to respond to growing demand for Claude but that the subscription service was not designed with third-party tool usage patterns in mind. He said the company would manage computing resources carefully and prioritize customers using its products and API.

After the change, it will still be possible to keep using Claude models in external agents such as OpenClaw. But users will have to go through the pay-as-you-go "Extra Usage" plan or the API.

Unlike the existing subscription plans that allowed unlimited use within a certain limit, the API charges based on the number of tokens used.

Anthropic said its own tools such as Claude Code and Cowork are optimized to raise the "prompt cache hit rate" that reuses previously processed text. Cherny said third-party services are not optimized in that way, making it difficult to sustain.

To reduce the impact, Anthropic will provide existing subscribers with a one-time credit worth the same amount as their monthly flat fee until April 17. Users can also receive up to a 30 percent discount if they pre-purchase Extra Usage packages.

Peter Steinberger (피터 스타인버거), an OpenClaw developer who recently joined OpenAI, pointed to what he described as Anthropic adding OpenClaw's popular features to its own Claude Code and then moving to block open source.