Anthropic officially bans third-party subscription authentication
Anthropic, the developer of the AI 'Claude,' updated its documentation on February 19, 2026, to clarify that using OAuth tokens with third-party tools violates their terms of use. Developers who used these credentials to access external tools have reported that their workflows have been interrupted.
Legal and compliance - Claude Code Docs
Anthropic Bans Claude Subscription OAuth in Third-Party Apps
https://winbuzzer.com/2026/02/19/anthropic-bans-claude-subscription-oauth-in-third-party-apps-xcxwbn/
Anthropic clarified that the OAuth tokens obtained with the Free, Pro, and Max plans are for use exclusively with Claude Code and Claude.ai, and that use of these tokens with any other products, tools, or services, including the Agent SDK, is unauthorized and a violation of its consumer terms of service.
As a result, developers building products or services that integrate with Claude's features, including the Agent SDK, must use API key authentication through the Claude Console or a supported cloud provider. Anthropic states that it 'does not permit third parties to provide Claude.ai logins or use Free, Pro, or Max plan credentials to route requests on your behalf.'
'Developers who relied on subscription-based authentication for external tools or automated workflows reported that their workflows were disrupted, with no prior notice,' technology media outlet WinBuzzer reported. 'The use of subscriptions in third-party apps has been in a gray area, but now it's been clarified.'
Anthropic's Thariq Shihipar later apologized, saying the documentation update was confusing, and advised users building businesses using the Agent SDK to use API keys instead of OAuth tokens.
We want to encourage local development and experimentation with the Agent SDK and claude -p.
— Thariq (@trq212) February 18, 2026
If you're building a business on top of the Agent SDK, you should use an API key instead. We'll make sure that's clearer in our docs.
Anthropic began enforcing measures against unauthorized access to its AI models around January 2026, implementing technical safeguards to block third-party coding workflow automation tools called 'Harness' from impersonating official clients.
Related Posts:
in AI, Posted by log1p_kr