Hermes is Nous Research's open-source autonomous agent: it lives on your server, keeps persistent memory, runs scheduled work, and reaches you over Telegram, Discord, or Slack. Its model section takes any OpenAI-compatible endpoint, so AIgateway drops in as the brain — and the agent can run on any of our 1000+ models without touching the agent itself.
One line on Linux / macOS / WSL2 (also Android via Termux). Reload your shell, then run setup; `hermes` on its own opens a chat once configured.
Hermes stores settings in ~/.hermes/config.yaml. Set provider: custom and Hermes calls base_url directly with your key. Set default explicitly — Hermes only auto-detects a model when the endpoint serves exactly one, and AIgateway serves the whole catalog.
Prefer not to hand-edit YAML? Run `hermes model`, choose the custom-endpoint option, and enter the base URL (https://api.aigateway.sh/v1), your key, and the model slug (moonshot/kimi-k2.7-code).
Change default to any provider/slug from the AIgateway catalog — anthropic/claude-opus-4.7, openai/gpt-5.4, moonshot/kimi-k2.7-code. Same agent, same memory, different model.