Anthropic, the developer behind the Claude large language model, has recently clarified its Consumer Terms of Service, specifically addressing the use of its services with third-party autonomous AI agents such as OpenClaw. This update, effective February 18, 2026, aims to regulate how users authenticate and interact with Anthropic’s models through external applications.
OpenClaw, an open-source AI agent, has garnered significant attention for its ability to extend generative AI capabilities beyond chatbots, enabling actions like running commands, interacting with APIs, and managing workflows on a user’s behalf. Originally launched as “Clawdbot” in November 2025, it underwent rebrands to “Moltbot” and finally “OpenClaw” following trademark complaints from Anthropic, according to multiple sources.
The core of Anthropic’s policy update centers on authentication methods. The company’s revised terms explicitly state that “Using OAuth tokens obtained through Claude Free, Pro, or Max accounts in any other product, tool, or service — including the Agent SDK — is not permitted and constitutes a violation of the Consumer Terms of Service.” This means users who were leveraging their Claude subscription OAuth tokens to power OpenClaw and similar tools are now in non-compliance.
According to reports and community discussions, this policy shift led to some users experiencing disabled Claude accounts. However, Anthropic later provided a clarification, stating that “nothing changes around how customers have been using their account and Anthropic will not be canceling accounts.” The company framed the update as a “clarification of existing language in our docs to make it consistent across pages.”
The critical takeaway for users is the distinction between OAuth tokens and API keys. While OAuth authentication from consumer subscriptions is now prohibited for third-party integrations, Anthropic confirmed that the use of official API keys for connecting with OpenClaw and other tools remains permissible. This mandates a switch for many developers who previously used the OAuth method, often to save on token usage costs associated with the flat-rate consumer subscriptions.
Community speculation suggests that Anthropic’s enforcement is largely a business decision. Many believe that heavy usage of Claude’s models through OpenClaw via fixed-price consumer subscriptions was economically unfavorable for Anthropic, as these users might consume significant tokens without contributing to data for model improvement through official channels. Peter Steinberger, the developer behind OpenClaw, announced on February 14, 2026, that he would be joining OpenAI and the OpenClaw project would be moved to an open-source foundation.
The widespread adoption of OpenClaw also highlighted security concerns. Experts noted that OpenClaw’s deep access to local systems for task execution, combined with instances of misconfigured or publicly exposed deployments, could pose security risks, including susceptibility to prompt injection attacks. Users are advised to transition to API keys for continued, compliant use of Anthropic’s models with OpenClaw, or explore local models for enhanced privacy and cost control.