Hermes Agent Third-Party API Quick Start: Install, Configure, and Verify

Install Hermes Agent, configure a third-party OpenAI-compatible API, choose a model, and verify that terminal chat works correctly.

Hermes Agent is an AI agent tool that can run locally or on a server. It works from the terminal and can also be connected to chat platforms such as Feishu or Telegram. Because Hermes supports OpenAI-compatible APIs, you can point it to a third-party gateway such as NixAPI and use GPT, Claude, Gemini, DeepSeek, MiniMax, and other models with one configuration.

This guide follows the shortest path: install Hermes Agent, enter the Base URL and API Key, choose a model, then verify that the terminal chat works.

Prerequisites

Before you start, prepare:

A macOS, Linux, or Windows machine
A NixAPI API Key
Terminal or PowerShell access

If you do not have a key yet, create an account from the NixAPI homepage, then open the NixAPI console and create an API Key. NixAPI is OpenAI-compatible, so in Hermes Agent you can choose OpenAI or OpenAI Compatible as the provider type.

NixAPI currently has a limited-time top-up exchange rate: RMB 0.80 = USD 1.00. If you plan to run longer agent tasks, check your balance before testing.

Step 1: Install Hermes Agent

macOS / Linux

Open your terminal and run the installer:

curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash

After installation, reload your shell configuration:

source ~/.zshrc

If you use bash, run:

source ~/.bashrc

Windows

Open PowerShell as Administrator and run:

irm https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.ps1 | iex

If Windows shows a security prompt, allow the script to continue. The first installation may take a few minutes depending on your network and local dependencies.

Step 2: Configure NixAPI as the Model Provider

Run:

hermes config

In the setup wizard, find the AI provider settings and fill in:

Field	Value
Provider type	`OpenAI` or `OpenAI Compatible`
API Key	Your NixAPI API Key
Base URL	`https://nixapi.com/v1`

The Base URL must include /v1:

https://nixapi.com/v1

If the wizard does not expose a Base URL field, edit the configuration file directly:

macOS / Linux: ~/.hermes/config.yaml
Windows: %USERPROFILE%\.hermes\config.yaml

Example:

provider:
  type: openai
  api_key: nix-your-api-key
  base_url: https://nixapi.com/v1
  model: claude-sonnet-4-20250514

Replace nix-your-api-key with your real key. For the model name, copy the exact model ID from the NixAPI supported models page.

Step 3: Choose a Model and Start Chatting

After saving the provider configuration, run:

hermes model

You can start with one of these model IDs:

claude-sonnet-4-20250514
claude-opus-4-7
gpt-5.4
gemini-3-pro-preview
deepseek-v4
minimax-m3

Model IDs must match the platform exactly. Pricing varies by model, so check the NixAPI model pricing page before long-running tests.

Start Hermes:

hermes

Then send a simple prompt:

Turn today's tasks into a prioritized checklist.

If Hermes returns an answer, the installation, API Key, Base URL, and model configuration are working.

Step 4: Optional Chat Platform Gateway

Hermes Agent can also connect to external chat platforms through its gateway. For Feishu, open a new terminal and run:

hermes gateway setup

Choose feishu, then enter the App ID and App Secret from your Feishu developer app.

On Windows, if you see a missing lark-oapi dependency error, install it into the virtual environment used by Hermes instead of your system Python:

<hermes-venv-path>/bin/pip install lark-oapi

Troubleshooting

`401 Unauthorized`

Check whether the API Key was copied completely, whether it has leading or trailing spaces, and whether the NixAPI account is active. If needed, create a fresh test key from the NixAPI console.

`404 Not Found`

The Base URL is usually wrong. For Hermes Agent, use:

https://nixapi.com/v1

Do not omit /v1.

`model not found`

The model name does not match a supported model ID. Copy it from the NixAPI supported models page.

How do I switch models?

Use /model in the Hermes chat, run:

hermes model

or edit the model field in ~/.hermes/config.yaml.

Summary

Hermes Agent runs the local agent workflow, while NixAPI provides a unified OpenAI-compatible model endpoint. Once configured, you can reuse the same Base URL and API Key and switch models by changing only the model ID.

Hermes Agent Third-Party API Quick Start: Install, Configure, and Verify Models