Running a Hermes agent with Claude Code subscription billing — no API key, no second meter — just became real: NousResearch shipped an official plugin (claude-subscription-directsdk, listed 20 September, v0.3.0) that runs Hermes turns through the unmodified official Claude Code CLI on your Claude Pro or Max plan. Same brain, one door, one bill. I’ve wired it into my Agent OS already — here’s the setup, the measured usage numbers nobody else will show you, and the account-safety caution this topic genuinely deserves.
Short answer
- The plugin: claude-subscription-directsdk (v0.3.0, official Hermes catalog, MIT, experimental) — Hermes turns on your Claude Pro/Max subscription via the official claude CLI. No API key, no per-token bill.
- Setup is ~3 minutes: Claude Code installed + logged in → Hermes updated to 0.21.4+ → one plugin install command → create a profile → test.
- Measured metering (plugin’s own Sept tests): bills at the claude -p / Agent SDK rate — about 1.7× the interactive TUI’s rate, so plan for ~60% of Claude Code’s throughput per 5-hour window.
- Fable 5.1 warning: on Pro it draws paid usage credits from the first request (Anthropic’s plan rule) — stick to Opus/Sonnet unless you accept that.
What shipped: Hermes agent with Claude Code subscription billing
The old situation was the gym-membership joke from my video: you paid for Claude, but using it inside an agent meant paying again — API key, separate balance, separate meter. Same brain, two doors, two bills. The plugin closes the second door: Hermes talks to the plugin, the plugin drives the official claude CLI (the Agent SDK path), and your existing Claude Code login does the rest. Hermes keeps everything that makes it Hermes — its agent loop, tools, memory, approvals and compaction — while Opus 5, Sonnet 5 or Haiku answer on the plan you already pay for.
It’s official on the Hermes side in the strong sense: published in the plugin catalog by NousResearch, pinned to a reviewed commit, updated only through reviewed re-pins — and refreshingly honest, labelling itself experimental and disclosing exactly how it bills. In my Agent OS it simply shows up as another profile: I made one called Claude Opus, typed “working?”, and it was — then ran it through Oracle, Muse and Astros like any other brain.
The 3-minute setup
| Step | Command / action |
|---|---|
| 1. Claude Code installed + logged in | npm install -g @anthropic-ai/claude-code then claude auth login |
| 2. Update Hermes | Version 0.21.4 or newer — older versions refuse the plugin cleanly rather than half-working |
| 3. Install the plugin | hermes plugins install claude-subscription-directsdk |
| 4. Create the profile | hermes profile create claude-opus (or pick the provider via hermes model) |
| 5. Test it | Send a message — the CLI does the talking; no API key is requested anywhere |
The plugin checks for the claude CLI at every seam — load, model selection, pickers, every request — and stops with a clear install hint instead of a traceback when it’s missing. The old way was twenty minutes of key-wrangling; this is genuinely about three. The deeper mechanics — what the CLI actually does per request — live in the CLI pipeline page, and the login details in the auth guide.
🔥 Want this set up without the guesswork? Wiring subscription-powered Claude into a full Agent OS 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 numbers nobody shows: metering, models, limits
The plugin’s README publishes real measurements, and they deserve summarising straight. Usage draws at the claude -p / Agent SDK rate — identical to running the CLI in print mode, no Hermes-specific penalty — but interactive Claude Code is metered at roughly 0.58× that rate, so budget for about 60% of the TUI’s throughput per five-hour window (their list-equivalent example: ~$6.5 vs ~$11 of Sonnet 5 on Pro). Partial consolation: on identical tasks Hermes sent ~0.6× the tokens native Claude Code did. Model routes: Sonnet 5, Opus 5, Opus 4.8 and Fable 5.1 at 1M context, Haiku 4.5 at 200K — with the crucial plan rule that Fable 5.1 on Pro bills to usage credits from the first request (Max plans include it up to half your weekly limits). That’s Anthropic’s documented rule, surfaced verbatim — and exactly why I said in the video I’d stick to Opus unless you accept the cost.
Account-safety honesty, up front: this plugin is “Official” in the Hermes catalog — reviewed and pinned by NousResearch — which is not the same as Anthropic endorsement. Commenters on my video raised real concerns, including claims of past account bans for routing Claude subscriptions through third-party harnesses, and my advice is exactly what I replied there: check Anthropic’s current terms of service before connecting your account. The plugin’s own disclosure is equally plain: every turn bills against your subscription’s Agent SDK allowance, and it inherits whatever account the claude CLI is logged into. Its authors label it experimental. Read the terms, read the README, decide informed — as of 23 September 2026.
Weigh it against the alternatives honestly: Astra rides its subscription inside Hermes too, and the API route remains the predictable-metering lane — this plugin’s pitch is specifically for people already paying for Claude Code who want one bill.
The bottom line on Hermes agent with Claude Code subscription
The hermes agent with claude code subscription setup delivers exactly what it promises — Claude’s best models inside a real agent on the plan you already pay, three minutes of setup, no API key — with two eyes-open realities: the ~60%-of-TUI metering rate, and the terms-of-service question you should answer for yourself before connecting the account you care about. Powerful, official-on-the-Hermes-side, experimental by its own label — go in informed.
FAQ: hermes agent with claude code subscription
Can I use my Claude Code subscription with Hermes agent?
Yes — the official Hermes plugin claude-subscription-directsdk (v0.3.0) runs Hermes turns through the claude CLI on your Pro/Max plan, no API key needed.
How long does setup take?
About three minutes if Claude Code is already installed: update Hermes to 0.21.4+, one plugin install command, create a profile, test.
How is usage billed?
Against your subscription’s Agent SDK allowance at the claude -p rate — roughly 1.7× the interactive TUI’s metering, so ~60% of its throughput per 5-hour window.
Which models work?
Sonnet 5, Opus 5, Opus 4.8, Fable 5.1 (1M context each) and Haiku 4.5 (200K) — but Fable on Pro draws paid usage credits from the first request.
Is this approved by Anthropic?
“Official” means official in the Hermes catalog — not Anthropic endorsement. Community members have raised account-safety concerns; check Anthropic’s current terms before connecting.
Do I still need an API key for anything?
Not for this route — the plugin sends no Hermes API key and actively rejects inherited API-key and custom-endpoint overrides.
Next step: if you want Claude working inside your agent on the plan you already pay 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 Claude CLI: The Pipeline Behind The Plugin
- Hermes Agent Claude Code Setup Is Actually Wild
Last updated September 2026. This is the living guide to hermes agent with claude code subscription — it gets updated as the tools change.
