What is the best subscription for Hermes Agent?
For an always-on Hermes Agent, the best subscription is a fixed-price API such as Standard Compute (from $19 a month); pick Nous Portal instead if you want its hosted tools, or a ChatGPT plan you already pay for if Codex models are enough. Hermes itself is free, so each option only pays for model calls.
What each option costs per month
| Option | Price | Always-on month | Limit |
|---|---|---|---|
Free trial model (nous/welcome) | $0 | $0 | Setup chat only, once per OS user |
| Local model (llama.cpp) | $0 plus hardware | $0 per token | Only models that fit your GPU |
| Nous Portal Free | $0 | $0 | Free models only |
| Nous Portal Plus, Super, Ultra | $20, $100, $200 | Plan price | $22, $110 or $220 of credits, also spent on tools and hosting |
| ChatGPT plan (Codex sign-in) | $20 Plus, from $100 Pro | Plan price | Codex models only, within the plan's Codex allowance |
| OpenRouter | Pay as you go | $6 to $189, plus 5.5% | None: the bill grows with use |
| Vendor API (OpenAI, Anthropic, DeepSeek) | Pay per token | $6 to $189 | None: the bill grows with use |
| Standard Compute | $19, $89, $249 | Plan price | $20, $95 or $269 compute budget, then HTTP 402 |
Always-on month: our median agent-day × 30 at list prices, from GPT-6 Luna to Claude Opus 5.5, measured on all agent traffic on our gateway (not Hermes alone), September 13 to October 12, 2026. Prices from Nous Portal, ChatGPT and OpenRouter, checked October 14, 2026.
Which one fits you
| If you | Pick |
|---|---|
| Are just trying Hermes | The free trial model, then Nous Portal Free or OpenRouter's free models |
| Have a GPU with enough memory | A local model: Hermes v0.22 installs llama.cpp and suggests models that fit |
| Already pay for ChatGPT Plus or Pro | ChatGPT sign-in. Hermes's docs don't say how its calls count against the plan, so watch the Codex allowance |
| Want web search, browser tools or Hermes Cloud hosting | Nous Portal, and budget its credits for the tools as well as the models |
| Use Hermes lightly and want model choice | OpenRouter or one vendor's API key |
| Run Hermes all day | A fixed monthly plan: Standard Compute $19 for an efficient model, $89 or more for Sonnet-class models |
Set up the one you pick
| Option | Setup in Hermes |
|---|---|
| Nous Portal | hermes setup --portal |
| ChatGPT plan | hermes model → ChatGPT or Codex Subscription |
| OpenRouter | OPENROUTER_API_KEY in ~/.hermes/.env |
| Local model (llama.cpp) | One click in the desktop app, or hermes model → Custom endpoint http://localhost:8080/v1 |
| Vendor API | hermes model, then the provider and your key |
| Standard Compute | standardcompute launch hermes, or Custom endpoint https://api.stdcmpt.com/v1 with model standardcompute |
From the Hermes provider docs and the v0.22.0 release notes.
The catch
Every fixed price here has a limit behind it. Portal credits can run out in days on a frontier model, and only $10 to $100 rolls over. A ChatGPT plan stops at its Codex allowance until the window resets. Standard Compute returns HTTP 402 at its compute budget until the plan renews or you upgrade.
Where Standard Compute fits
Standard Compute is an OpenAI-compatible API with plans from $19 a month for a $20 compute budget and no overage charges. With the model standardcompute, each request is routed to an efficient or a stronger model by how hard it is; an exact model ID pins one model.
FAQ
What is the best subscription for Hermes Agent?
For an agent that runs all day, a fixed-price API such as Standard Compute, from $19 a month. Nous Portal is the better pick for its hosted tools and cloud hosting, and a ChatGPT plan you already have covers Codex models.
How much does Nous Portal cost for Hermes Agent?
Free with free models only, then Plus $20, Super $100 and Ultra $200 a month, which buy $22, $110 and $220 of credits. Up to $10, $50 or $100 of unused credit rolls over. Credits pay for models, hosted tools and Hermes Cloud hosting.
Is Hermes Agent free?
Yes. Hermes Agent is open source under the MIT license, desktop apps included. You pay only for the models it calls, unless you run a local model or stay on free models.
Can I use my ChatGPT subscription with Hermes Agent?
Yes. Run hermes model and pick ChatGPT or Codex Subscription; Hermes then uses Codex models through your ChatGPT sign-in. Nous Portal also accepts an eligible ChatGPT plan for inference without a paid Nous subscription, with tools and hosting billed separately.
Can I use my Claude subscription with Hermes Agent?
Not as a flat plan. Hermes's Anthropic sign-in works only with Claude Max plus purchased extra usage, and it spends only that extra usage, at API rates; the base Max allowance is untouched. Claude Pro sign-in is not supported, so Pro users need an Anthropic API key.
Does Hermes Agent have a free model?
Since v0.22, a first desktop install with no provider gets the free nous/welcome model once per OS user, for the setup chat. For ongoing free use, Nous Portal Free and OpenRouter offer free models with rate limits, and local models through llama.cpp cost nothing per token.
What is the cheapest way to run Hermes Agent all day?
A local model if your GPU can run one well. Otherwise an efficient model such as GPT-6 Luna or DeepSeek V4.1 Flash, about $6 to $7 a month for a typical always-on agent at list prices, on its own API key or on Standard Compute's $19 plan.