Connect Any Model Provider

Hermes Agent works with more than twenty LLM providers. The agent, its skills, its memory and its workflows stay exactly the same — only the model changes. That single design choice protects you from lock-in: a price change, an outage or a better model elsewhere is a one-command switch.

The providers at a glance

Switching models: the interactive way

hermes model      # interactive picker — browse and select
hermes model      # inside a session: /model does the same

The picker writes your choice to ~/.hermes/config.yaml under model.default. Change it as often as you like — every later session uses the new model, and nothing else moves.

Credential pools: never dead-end mid-task

hermes auth add               # register another key for a provider
hermes auth list openrouter   # see the pool
hermes auth reset openrouter  # clear exhaustion flags

When one key runs dry, Hermes moves to the next in the pool automatically. For heavy daily use this is the difference between a workflow that stops and one that does not.

Which provider should you start with?

If you already pay for a coding subscription (Nous Portal, OpenAI Codex), start there — the OAuth logins reuse it. If you want maximum flexibility with one bill, OpenRouter is the common choice: a single key reaches frontier and budget models alike, which pairs well with Hermes's habit of letting you pick a cheaper model for chores and a stronger one for hard problems. Whatever you choose, hermes doctor will confirm the connection before your first task.

Rotating more than one provider

Mixed setups work exactly like single ones. Keep a frontier model as your default for hard problems, point chores at a budget model, and register a spare credential for whichever provider you lean on hardest — the pool absorbs the rate limit you will eventually hit. Nothing about that mix lives anywhere but your own config, which is the whole point: your agent stack should not change when a vendor's pricing page does.