Hermes Agent

Hermes Agent Integration

Use UnoRouter as the model provider for Hermes Agent. Point one OpenAI-compatible endpoint at UnoRouter and reach every model.

What is Hermes Agent?

Hermes Agent by Nous Research is an open-source, self-improving AI agent. It runs in the terminal, as a messaging gateway, and as an IDE integration, keeps durable memory on disk, and works with any OpenAI-compatible endpoint.

Core Features

Self-improving - Builds skills from experience and a persistent model of your preferences across sessions.
OpenAI-compatible - Routes every chat turn through one custom endpoint, so UnoRouter serves all models.
Multiple interfaces - A classic CLI, a modern TUI, messaging gateways, and IDE integration over ACP, all on one core.
Gateway and proxy - Exposes an OpenAI-compatible API server and a local subscription proxy for OAuth-backed providers.

Installation

Install with the script or via PyPI. The postinstall step pulls optional tooling and runs the setup wizard.

bash

# Install script (recommended)
curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash

# Or via PyPI
pip install hermes-agent
hermes postinstall

Windows requires WSL2

Hermes Agent runs under WSL2 on Windows. Install and configure it inside your WSL2 distribution, where the config lives at ~/.hermes/config.yaml.

Configuration

Edit ~/.hermes/config.yaml to route Hermes through UnoRouter. With provider set to custom, base_url takes precedence and every turn goes to UnoRouter:

~/.hermes/config.yaml

yaml

model:
  provider: "custom"
  base_url: "https://api.unorouter.ai/v1"
  default: "claude-opus-4-8"

Store the API key in ~/.hermes/.env (or set it in your shell). Hermes uses OPENAI_API_KEY for a custom endpoint:

bash

OPENAI_API_KEY=YOUR_API_KEY

Key Configuration Details

Field	Description
model.provider	Set to custom so Hermes calls your endpoint directly instead of a built-in provider.
model.base_url	The UnoRouter OpenAI-compatible base URL. Hermes verifies it against /v1/models on startup.
model.default	The default model id. Add more ids under a models list to populate the model picker.
OPENAI_API_KEY	Your UnoRouter API key from the dashboard, read from ~/.hermes/.env.

Pick a 64k+ context model

Hermes requires a model with at least 64,000 tokens of context. Smaller windows are rejected at startup because multi-step tool use needs the working memory.

CLI alternative

Prefer commands over editing the file? Hermes routes each value to the right file automatically:

bash

hermes config set model.provider custom
hermes config set model.base_url https://api.unorouter.ai/v1
hermes config set OPENAI_API_KEY YOUR_API_KEY

Run Hermes

Start a session, switch model or provider at any time, or diagnose your setup:

bash

# Interactive TUI (recommended)
hermes --tui

# Switch model or provider any time
hermes model

# Diagnose configuration
hermes doctor

Ready to get started?

Get your API key and start using UnoRouter with Hermes Agent.