Anthropic announced on April 4 that Claude Pro and Max subscribers now have to pay extra to use third-party AI agents such as OpenClaw, a popular automation tool. The change ends the practice of applying existing subscription limits to these integrations, a shift that redraws the line between direct service and external tooling. Anthropic cited a disproportionate resource load created by Claude's "brain" powering OpenClaw workflows.

The new model shifts third-party usage to a pay-as-you-go pricing structure. Subscriptions will still include access to Claude.ai, Claude Code, and Claude Cowork, but any interaction with external agents now incurs additional charges. Anthropic clarified that its core plans were not designed for heavy third-party tool usage and that resources are allocated deliberately to prioritize direct customers, a philosophy that mirrors the recent introduction of unified billing in its enterprise marketplace.

A company spokesperson explained, "We're working hard to meet the growing demand for Claude, and our subscriptions were not built for third-party tool usage." Users who object to the new terms can request a refund. Anthropic also offers discounts of up to 30% on extra-usage packages and a one-time credit equal to a month's subscription, concessions that acknowledge the sudden policy shift while maintaining the structural logic of separating infrastructure costs.

Anthropic will initially apply the policy only to OpenClaw and roll it out gradually to other third-party solutions. Existing users are encouraged to review their usage patterns and consider the optional discount bundles. The move signals a broader shift in how AI companies measure agent-driven workflows versus traditional chat interfaces. For more details on Claude's evolving feature set, see the recent macOS control addition for Claude Pro and Max.