Hermes Agent + OmniaKey: a custom OpenAI-compatible endpoint

Hermes Agent (from Nous Research) ships with built-in providers, but it's designed to talk to any OpenAI-compatible endpoint. That makes OmniaKey a clean fit: one custom endpoint, and the same key reaches Claude, GPT, and Gemini — you switch model by id, not by reconfiguring.

The fast path: hermes model

The quickest setup is the interactive picker:

hermes model

Choose Custom endpoint (self-hosted / VLLM / etc.), then enter:

• Base URL: https://api.omniakey.com/v1
• API key: your OmniaKey key
• Model: e.g. claude-opus-4-8

Or by hand: config.yaml

Prefer to edit it directly, or want a persistent multi-model setup? Put the same thing in ~/.hermes/config.yaml instead:

model:
  provider: custom
  base_url: https://api.omniakey.com/v1
  api_key: your-omniakey-api-key
  default: claude-opus-4-8
models:
  - claude-opus-4-8
  - gpt-5.5
  - gemini-3.1-pro-preview

provider: custom is what tells Hermes to call your endpoint directly with the key above, instead of one of its built-in providers. The models: list is what populates the /model picker — restart Hermes once after editing, and you can switch between claude-opus-4-8, gpt-5.5, and gemini-3.1-pro-preview without leaving the session.

One key, three families

Because OmniaKey routes by model id over the OpenAI-compatible surface, a single custom endpoint covers all three families — no second provider block, no juggling base URLs. Billing is per token against one prepaid balance, with no monthly plan. And the model id you set is the one that runs: no silent fallback to a cheaper "equivalent" that would reason about your codebase differently mid-task.

The coding agents guide covers the other tools.