Hermes Claude is a pairing with a plot twist: Claude used to work in Hermes, then it didn’t — the door closed — and as of 20 September 2026 an official NousResearch plugin reopened it, running Claude through your existing Claude Code subscription with no API key. I’ve already got Opus answering inside my Agent OS workflows. This is the pairing’s hub page: the history, what the plugin actually is, the five model routes with their real context windows, what works inside the loop, and the cautions the comment section rightly demanded.
Short answer
- The comeback: Claude in Hermes stopped working; the official claude-subscription-directsdk plugin (listed 20 Sept, v0.3.0, experimental) restores it via the official claude CLI on your Pro/Max plan.
- Five routes: Sonnet 5, Opus 5, Opus 4.8 and Fable 5.1 at 1M context; Haiku 4.5 at 200K — short names (opus, sonnet…) resolve automatically.
- Hermes stays Hermes: its own agent loop, tools, approvals, memory and compaction — Claude is the brain, not the body.
- Two cautions: metering runs at the Agent SDK rate (~60% of the TUI’s throughput), and “official” means Hermes-official — check Anthropic’s ToS yourself.
The Hermes Claude story: gone, then back
For months the answer to “can I run Claude in Hermes?” was a flat no — the previous route stopped working, and the workaround was paying separately for API access even if you already had a Claude subscription. Two doors, two bills, same brain. The new plugin ends that era: Hermes → plugin → the unmodified official Claude Code CLI → your subscription. No API key, no second account. In my Agent OS it’s now just a profile called Claude Opus sitting in the model dropdown next to everything else.
What makes this different from a hack is the engineering posture: catalog-listed and pinned to a reviewed commit, scrupulous about credentials (they stay with the official CLI — the plugin never opens or copies them), honest about being experimental, and publishing its own measured metering rather than pretending usage is free. The full anatomy lives across this cluster: the subscription setup, auth and the CLI pipeline.
What Hermes Claude gives you in practice
| Model route | Context window | Plan note |
|---|---|---|
| Sonnet 5 | 1,000,000 | The workhorse — metering examples below use it |
| Opus 5 | 1,000,000 | My default profile for heavy reasoning |
| Opus 4.8 | 1,000,000 | Kept in the picker for continuity |
| Fable 5.1 | 1,000,000 | ⚠ On Pro: paid usage credits from the first request; Max includes it up to 50% of weekly limits |
| Haiku 4.5 | 200,000 | The fast/cheap lane |
Inside the loop, it behaves like Hermes with a Claude brain should: reads files, runs commands, uses the web, checks its own work — and slots into every workflow I run, from Oracle research loops to Muse and Astros. Because Hermes owns the loop, all its disciplines still apply: /goal budgets, approvals, memory profiles, compaction. Claude supplies the thinking; Hermes supplies the hands and the rules.
🔥 Want this set up without the guesswork? A full Hermes + Claude Agent OS, set up properly 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 honest cautions
Metering first: the plugin’s own measurements show subscription draw at the claude -p / Agent SDK rate — about 1.7× the interactive TUI’s rate — so a Pro plan buys roughly 60% of the throughput per five-hour window that native Claude Code would give you (softened somewhat by Hermes sending ~0.6× the tokens on identical tasks). Model entitlements follow your plan: the Fable rule above is Anthropic’s own, and the picker labels credit-drawing routes rather than hiding them.
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.
My practical read: run it on Opus or Sonnet, watch your usage the first week (compare Hermes /usage against Claude Code /cost on the same task, as the README itself suggests), and treat this the way its authors do — a powerful experimental bridge, not settled infrastructure.
The bottom line on Hermes Claude
Hermes Claude went from dead pairing to the most interesting brain-swap in the agent world in one plugin release: Claude’s frontier models, 1M-context routes, inside a real agent, on the subscription you already pay — experimental label, metering maths and ToS homework included. If you’ve been running Hermes on anything else while missing Claude’s reasoning, the door’s open again. Walk through it with your eyes open.
FAQ: hermes claude
Does Claude work with Hermes agent now?
Yes — as of 20 September 2026, via the official claude-subscription-directsdk plugin, which runs Claude through your Claude Code subscription with no API key.
Why did Claude stop working in Hermes before?
The previous route was shut off — leaving API-key billing as the only path until this official plugin restored subscription access.
Which Claude models can Hermes use?
Sonnet 5, Opus 5, Opus 4.8 and Fable 5.1 at 1M context, plus Haiku 4.5 at 200K — with Fable drawing paid credits on Pro plans.
Does Hermes lose any features with Claude as the brain?
No — Hermes keeps its own loop, tools, approvals, memory and compaction; Claude replaces only the thinking.
Is this the same as Claude Code?
No — it uses Claude Code’s CLI as transport, but the agent is Hermes: different loop, different tools, different metering rate (~1.7× the TUI’s).
Is it safe for my Claude account?
That’s the open question — Hermes-official is not Anthropic-endorsed, commenters have claimed past bans for third-party routing, and my advice is to check Anthropic’s current terms before connecting.
Next step: if you want Claude thinking inside your Hermes workflows this week 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 Agent + Claude Code Subscription: No API Key
- Hermes-Claude Auth: How Login Works (No API Key)
- Hermes Agent Claude Code Setup Is Actually Wild
Last updated September 2026. This is the living guide to hermes claude — it gets updated as the tools change.