The Hermes-Claude auth story is the most reassuring part of the new subscription plugin — because the answer to “who holds my credentials?” is: not the plugin. Authentication belongs entirely to the official Claude Code CLI: you run claude auth login once, and the plugin never opens, copies, refreshes or prints those credential files — it just drives the CLI that already holds them. No API key exists anywhere in the flow. Here’s exactly how the auth chain works, the config options for separated setups, and the account-safety context you should weigh before logging in at all.
Short answer
- Auth lives in the official CLI:
claude auth login(browser + pasted code) — the plugin never opens, copies, refreshes or prints credential files. - No API key anywhere: none required, none sent — and the plugin actively REJECTS inherited API-key, custom-endpoint and cloud overrides, naming the conflicting variables without printing their values.
- Status checks are local:
claude auth statusreads the local credential store in ~0.3s and showscredentials: ✓ (Claude Pro). - Separated setups:
CLAUDE_CONFIG_DIR/CLAUDE_SUBSCRIPTION_DIRECTSDK_CONFIG_DIRpoint at a dedicated CLI-managed auth directory.
How hermes-claude auth actually works
The chain is deliberately boring: you authenticate the official Claude Code CLI exactly as Anthropic designed — claude auth login opens the browser, you paste the code, done. From then on, every Hermes request spawns the unmodified claude executable, which presents its own credentials to its own service. The plugin’s role is transport, not custody: per its documentation, it has no credentials of its own, and native authorization headers pass through memory directly to the upstream — never logged, never persisted.
Selection-time checks are equally tidy: choosing the provider first asks the CLI itself — installed? (PATH check, or the override variable), logged in? (claude auth status against the local store) — and if you’re logged out on a terminal, it starts claude auth login inline and re-checks. Nothing is written to Hermes config until those pass. The model list itself comes from the CLI’s initialize handshake with zero upstream requests — verified through the plugin’s own admission relay.
The guardrails: what the plugin refuses to do
| Guardrail | What it means for you |
|---|---|
| No API key sent, ever | This route can’t silently fall back to metered API billing |
| Rejects inherited overrides | Conflicting ANTHROPIC/API-key/custom-endpoint env vars abort the request — named, values unprinted |
| Credentials never touched | The plugin can’t leak what it never reads — files stay with the official CLI |
| Headers pass through memory | Not logged, not persisted; cancellation tears down the connection and listener |
For separated setups — say, a dedicated auth directory you manage independently — run CLAUDE_CONFIG_DIR=/path claude auth login once, then export CLAUDE_SUBSCRIPTION_DIRECTSDK_CONFIG_DIR=/path for the plugin. An inherited CLAUDE_CONFIG_DIR also works, and CLAUDE_SUBSCRIPTION_DIRECTSDK_COMMAND selects a claude binary outside PATH. Full CLI mechanics live in the pipeline page.
🔥 Want this set up without the guesswork? A safely-configured Hermes + Claude setup, end to end is exactly the kind of thing we set up together inside the AI Profit Boardroom — 3,700+ members, four live calls a week, daily tutorials, done-for-you templates and a 30-day roadmap. Prefer 1-on-1 help? Book a free SEO strategy session and we’ll map it out for your business.
The part auth can’t protect you from
Everything above is about credential hygiene, and it’s genuinely well done. What it cannot answer is the policy question: whose rules govern using your subscription through a third-party harness at all. The plugin’s own disclosure is direct — it “inherits whatever account claude is logged into” and bills that account’s Agent SDK allowance — which means the account you log in with is the account taking whatever risk exists.
Account-safety honesty: this plugin is “Official” in the Hermes catalog — reviewed and pinned by NousResearch — which is not Anthropic endorsement. Commenters on my video raised real concerns, including claims of past account bans for routing Claude subscriptions through third-party harnesses; my advice is what I replied there: check Anthropic’s current terms of service before connecting your account. The plugin’s own disclosure: every turn bills your subscription’s Agent SDK allowance, and it inherits whatever account the claude CLI is logged into. Experimental by its authors’ own label — as of 23 September 2026.
Practical posture while the dust settles: read Anthropic’s current terms yourself (they, not any plugin README, govern your account), keep the metering realities in mind so usage surprises don’t compound policy questions, and remember that disabling extra usage in your Claude account settings caps overage exposure — the README points that out itself.
The bottom line on hermes-claude auth
Hermes-Claude auth is the right architecture done carefully: the official CLI owns login, the plugin owns nothing, no API key exists to leak, and conflicting configurations fail loudly instead of silently. The remaining question isn’t technical — it’s whose terms you’re operating under, and only Anthropic’s current ToS answers that. Clean hygiene, open policy question: log in accordingly.
FAQ: hermes claude auth
How does authentication work between Hermes and Claude?
Entirely through the official CLI: claude auth login once; the plugin drives that CLI and never opens, copies, refreshes or prints its credential files.
Do I need an API key for Hermes-Claude?
No — none is required or sent, and the plugin rejects inherited API-key or custom-endpoint variables rather than silently using them.
How do I check I’m logged in?
claude auth status — a ~0.3s local check showing credentials ✓ with your plan; the provider runs it before anything is saved.
Can I use a separate auth directory?
Yes — authenticate with CLAUDE_CONFIG_DIR pointed at it, then set CLAUDE_SUBSCRIPTION_DIRECTSDK_CONFIG_DIR for the plugin.
Which account gets billed?
Whichever account the claude CLI is logged into — the plugin’s disclosure is explicit that it inherits that account and bills its Agent SDK allowance.
Is my account safe using this?
Credential hygiene is excellent; the policy question is Anthropic’s terms — commenters have claimed past bans for third-party routing, so read the current ToS before connecting an account you care about.
Next step: if you want your auth and account questions settled before you build working for you this week, join the AI Profit Boardroom for the full walkthroughs and live help — or book a free SEO strategy session and I’ll point you at the fastest path for your situation.
About Julian Goldie: SEO agency owner with 10+ years in SEO, 394K+ subscribers on YouTube, a 100% job-success score on Upwork, 75K+ members across his communities, and author of a best-selling SEO book. He runs the AI Profit Boardroom community and offers a free SEO strategy session.
Related reading
- Hermes Claude: The Official Plugin Brings It Back
- Hermes Agent + Claude Code Subscription: No API Key
- Hermes Claude CLI: The Pipeline Behind The Plugin
Last updated September 2026. This is the living guide to hermes claude auth — it gets updated as the tools change.