Pick a provider and give it a key

A provider is the language model Hermes talks to when it thinks. Hermes supports a dozen of them. For your first setup, ignore eleven and pick one of two.

Nous Portal is the default, made by the same team that builds Hermes. The setup is a single sign-on flow — no API key to manage, no billing to configure. If you want zero friction, start here.

OpenRouter is our other pick. It's one account that gives you access to hundreds of models from different providers, with pay-as-you-go pricing. You can use Claude, GPT-4, Llama, DeepSeek, and a lot more, all through one API key. Slightly more setup than Nous Portal, much more flexibility.

You can change your mind later with one command. There is no cost to picking "wrong."

Option A: Nous Portal (easiest)

Run the model picker:

hermes model

You'll see a list of providers. Pick Nous Portal and follow the prompts. It'll walk you through signing in and configuring the connection. When it's done, you're ready.

That's it. Skip to the checkpoint at the bottom of this page.

Option B: OpenRouter

Head over to openrouter.ai and create a free account. Once you're logged in, go to the Keys page and create a new API key. Copy it.

Now tell Hermes about it:

hermes config set OPENROUTER_API_KEY your_key_here

Then run the model picker and choose OpenRouter:

hermes model

Pick a model from the list. If you're not sure which, anthropic/claude-sonnet-4 or openai/gpt-4.1 are good defaults — they're smart enough to handle most agent work without being expensive. You can change models later with /model inside chat.

Add some credit

OpenRouter charges pay-as-you-go. Load five or ten dollars into your account before you start playing — it lasts a long time for personal use, and it saves you the embarrassment of hitting an empty-wallet error mid-conversation.

Other providers

If you have an OpenAI, Anthropic, DeepSeek, Kimi, or MiniMax key already, Hermes supports all of them. Run hermes model, pick the provider, and paste your key when prompted. The flow is similar to OpenRouter.

If you're running a local model with vLLM, SGLang, or Ollama, use the "custom endpoint" option in the model picker and point Hermes at your local URL.

Sanity check

Checkpoint — Run hermes model one more time and confirm your chosen provider has a checkmark or "active" marker next to it. If it does, Hermes knows how to talk to a language model.

If something went wrong

API key not set error when you try to chat. Run hermes model again and walk through the flow. If you're using OpenRouter, double-check that you ran hermes config set OPENROUTER_API_KEY <key> with the actual key, not the literal word your_key_here.

Model list is empty. You might be behind a firewall blocking the provider list endpoint. Try a different network, or fall back to OpenRouter — its endpoint is usually reachable.

Ready? Next step: actually talk to Hermes and watch it do real work.