Anthropic to bill Claude users separately for OpenClaw usage

Anthropic, the company behind the AI platform Claude, has recently announced that it will begin charging users separately for the use of OpenClaw, an open-source tool that allows AI models to control a user's computer and perform tasks. This move marks a shift from the previous subscription-based pricing model to a usage-based system, where users may have to pay extra for OpenClaw usage.

The decision to implement this new pricing structure has been communicated to subscribers through email notifications, as reported by users on Reddit and Hacker News. While OpenClaw usage will now be billed separately, users can still access third-party harnesses through a "pay-as-you-go" option. To help users adapt to this change, Anthropic is offering a one-time credit equal to the monthly subscription fee, along with discounts of up to 30% on prepaid usage.

In a statement, Anthropic explained that the tools provided by OpenClaw put a significant strain on their systems. The company emphasized the need to manage capacity carefully and prioritize customers using their core products. This shift in pricing is a direct response to the challenges posed by OpenClaw's usage patterns, which differ significantly from traditional chat-based interactions.

OpenClaw is an agentic open-source tool that enables AI models to control a user's computer and carry out tasks such as sending messages, browsing websites, managing files, and executing commands through a chat interface. The tool connects to various apps and services, allowing the AI model to interact with them directly. However, this capability has led to some controversy, as evidenced by Google's recent decision to restrict access to its Antigravity coding platform for some users who accessed it via OpenClaw in February 2026.

Google's move affected even paid subscribers and was linked to the use of third-party tools in ways that violated terms of service. The company stated that it acted after detecting a "massive increase" in malicious usage that degraded service quality for other users. This incident highlights the potential issues that can arise when AI agents are used to interact with external systems, particularly when they are not designed with such integrations in mind.

The move to usage-based pricing for OpenClaw is not unique to Anthropic. Companies are increasingly adopting this model for AI agents, as they fundamentally change cost dynamics. Unlike traditional chat use, agents execute multi-step workflows, call tools, and generate far higher token volumes per task. This means that costs scale with behavior, not just usage. Even simple requests can expand into thousands of tokens, and at scale, these interactions can rapidly drive up infrastructure costs. This creates a mismatch with flat subscriptions, which assume predictable usage patterns.

In light of these challenges, Anthropic's decision to bill users separately for OpenClaw usage is an attempt to better align pricing with the actual resource consumption and strain on their systems. By offering discounts and credits, the company aims to make this transition as smooth as possible for its users. As AI agents continue to evolve and become more integrated into various platforms, it will be interesting to see how other companies adapt their pricing models to accommodate these changing dynamics.