What is Anthropic changing about OpenClaw use?

Anthropic limits Claude subscriptions from third-party OpenClaw usage

Anthropic has introduced a new policy that restricts how Claude subscriptions can be used with third-party AI agent tools -- specifically OpenClaw. The change is intended to better manage capacity and comes after a period of heightened attention on OpenClaw's rapid growth and security concerns.

Under the updated terms described in reporting, Claude Pro and Max subscribers will no longer be able to use their flat-rate subscription to run Claude through OpenClaw. Instead, using Claude models via OpenClaw will require additional payment or separate usage-based access starting on April 4 (with a stated start time in Pacific time for one of the policy descriptions).

The move shifts OpenClaw usage from being effectively included in subscription coverage to becoming an additional billable path. In practical terms, this means the total cost of using Claude to power agent workflows through an external framework will likely increase for developers and teams that relied on OpenClaw as a way to operationalize coding and task automation.

The policy change also reinforces a theme emerging across the AI-agent ecosystem: platforms are trying to distinguish between first-party integrations and third-party tooling that may drive variable compute demand. By limiting subscription inclusions, Anthropic can more directly meter usage and control infrastructure costs.

For end users, the impact is mostly financial and operational: existing automations that previously assumed "subscription covers everything" may need to be reconfigured around new pricing or access methods. For the agent tooling ecosystem, it signals that third-party frameworks could increasingly be subject to tighter monetization and access controls from the underlying model providers.