Not safe for Oauth use with CC?

[The1Ice9]

After a review the wrap claude / proxy is a ban risk, no?

"
2. ToS — routing your subscription through the proxy is a ban risk. Don't.

I checked it now, and the picture changed under your feet in January 2026. The current Anthropic Consumer ToS language:

▎ "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."

Anthropic's own support docs: OAuth auth is "exclusively for ordinary use of Claude Code and other native Anthropic applications," and third parties "cannot route requests through Pro or Max plan credentials."

This is a new, distinct bright line from the "seat rule" in your memory. The seat rule is about who triggers a call; this is about which process holds and forwards the token. headroom wrap claude sets ANTHROPIC_BASE_URL so your OAuth bearer transits the Headroom proxy before reaching Anthropic — the token is being handled by "another tool/service," which is exactly the now-prohibited pattern. And enforcement is live and unannounced: blocking began Jan 9, 2026, docs clarified Feb 19, and real bans have been issued — including for a passive Claude Code usage-tracker app, so the line is "did the token touch third-party code," not intent. (It's also mechanically fragile: ANTHROPIC_BASE_URL only redirects inference; ~30 OAuth/admin endpoints are hardcoded to api.anthropic.com.)

So the earlier "verify, probably respects the seat-rule" was wrong given the Jan-2026 clause — corrected.

"