Anthropic has officially updated its acceptable use policy to prohibit using OAuth tokens from Free, Pro, and Max consumer subscriptions in any third-party product, tool, or service. This explicitly forbids routing requests through these plan credentials or offering Claude.ai login externally. The rule targets developers building applications with the Agent SDK or similar integrations.

Instead, the company mandates that developers building external products must use API key authentication obtained through the Claude Console or supported cloud providers like AWS Bedrock and Google Vertex. The policy draws a clear line: OAuth tokens are for direct end-user access to Claude.ai and Claude Code only, while API keys are for programmatic, server-side integration.

Enforcement of these restrictions may occur without prior notice. This move formalizes a key boundary for the ecosystem, preventing the indirect scaling of consumer-tier access and ensuring third-party integrations use the intended, auditable authentication method designed for commercial development.