Claude OAuth login / clarify ToS risk vs. Anthropic's stated policy

Thanks for jcode, I'm kinda happy to found it so far.

Quick flag: jcode login --provider claude uses Pro/Max subscription OAuth. Anthropic's compliance docs currently say those tokens are only for Claude Code and Claude.ai, explicitly including the Agent SDK as out of scope:

"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."

They've apparently been enforcing this since January 2026, and OpenCode dropped its bundled Claude OAuth plugin after a legal request. Same question came up on the Pi project:

https://github.com/earendil-works/pi/discussions/1999

Questions:

• Is this flow believed to be fine somehow, or the same pattern being enforced against elsewhere?
• Any known risk of it getting blocked like OpenCode's?
• Worth a heads-up in the README/CLI near that login command?

Just want the actual picture here, everything in this space seems to shift under our feet week to week.

Start by reviewing the implementation behind jcode login --provider claude, then compare its OAuth use with Anthropic's quoted policy and the linked Pi discussion. Confirm whether a project-side change is needed; if so, document the supported scope or risk near the README or CLI login guidance.