Nous Research agent

Hermes: connect a cheap AI API

Hermes takes any OpenAI-compatible backend, so it's easy to keep it on cheap models for routine work and switch to a flagship for the hard parts — all on one key and one balance.

Hermes

Self-improving agent by Nous Research (CLI)

Create an API key on the Keys page and copy it.

Install Hermes Agent (macOS / Linux / WSL2):

bash

curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash

Windows PowerShell: iex (irm https://hermes-agent.nousresearch.com/install.ps1)

Point Hermes at cheapai.io with a custom provider (OpenAI-compatible):

bash

hermes config set model.provider custom
hermes config set model.base_url https://cheapai.io/v1
hermes config set model.api_key YOUR_API_KEY
hermes config set model.model claude-opus-4-8

Or edit config.yaml directly:

yaml

model:
  provider: custom
  base_url: https://cheapai.io/v1
  api_key: YOUR_API_KEY
  model: claude-opus-4-8

Run hermes. The model is any slug from our catalog (e.g. claude-opus-4-8, gpt-5.6-sol, gemini-3-pro) without a vendor prefix — swap it anytime via hermes model.

Hermes uses the OpenAI chat/completions format, so any text model from the catalog works. The API key resolves from model.api_key or the OPENAI_API_KEY env var.

Replace YOUR_API_KEY with a real key from the Keys section.

Hermes — common questions

Does Hermes need its own key?

No, any key from the dashboard works. A separate one only helps if you want this agent's spend as its own line in the logs.

Get a key and plug it into Hermes

Claude, GPT, Gemini and Grok on one key · pay in ₽, USDT or by card · no VPN

Get API key

Guides for other tools

OpenCode Claude Code Codex Cursor Kilo Code Cline / Roo Zed Continue