Anthropic clarifies ban on third-party tool access to Claude subscriptions

Anthropic has revised its Consumer Terms of Service to explicitly prohibit third-party tools from accessing Claude subscriptions, formalizing a longstanding policy designed to protect its revenue model. The update specifically bans using OAuth tokens from Claude Free, Pro, or Max accounts with unauthorized interfaces.

The policy targets third-party harnesses—wrapper tools like OpenCode or Cursor that enhance user interaction with Claude's AI models. These tools circumvented Anthropic's pricing structure by allowing subscribers to feed prompts through cheaper subscription tiers instead of higher-cost API access. This practice, termed "token arbitrage," exploited Anthropic's subscription model where monthly fees function as de facto subsidies for heavy users.

Legal language in Section 3.7 now unequivocally states: "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." While this restriction existed since February 2024, Anthropic's January enforcement actions against "spoofing" tools prompted the clarification. Engineer Thariq Shihipar noted third-party tools caused operational issues by generating atypical traffic patterns without diagnostic telemetry.

The impact is immediate: OpenCode removed Claude support following "anthropic legal requests," and affected developers must now choose between official interfaces or pay-as-you-go API access. This move underscores frontier AI companies' struggle against disintermediation, where third-party tools threaten core revenue streams despite billion-dollar model investments. Competitor OpenAI openly permits similar integrations, highlighting divergent business approaches in the AI ecosystem.

Users violating the policy risk account suspension, while companies building harnesses face legal pressure to disable Claude integration—a significant constraint on third-party AI tool development.