Hermes Agent

Hermes Agent is a terminal AI coding tool. By configuring it, you can connect to AI models on the Unisound Maas Platform through pay-as-you-go and U Plan.

1. Preparation

Get configuration credentials

Two usage methods are supported, but the way to obtain the corresponding credentials differs.

Usage
Description
How to get it (the API Key below is an example)
Pay-as-you-go API calls
Billed based on actual usage, suitable for light usage
  • Base URL OpenAI-compatible protocol: https://maas-api.unisound.com/v1
  • API Key
  • Format: sk-xxxxx

Go to API Key Management to create an API Key

U Plan
Fixed subscription pricing with usage quotas based on the selected plan
  • Base URL OpenAI-compatible protocol: https://maas-api.unisound.com/v1
  • API Key
  • Format: tp-xxxxx

After subscribing successfully, go to Subscription Management to get the dedicated API Key

2. Install Hermes Agent

Hermes Agent supports Linux, macOS, and WSL2 (Windows). For more details, see the Hermes Agent official docs.

  • Linux / macOS does not require extra setup; the default environment is enough.
  • On Windows, refer to Install WSL on Windows to install WSL2 first, then run the commands below inside WSL2.

Installation command:

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

After installation, reload the shell environment:

Bash
source ~/.bashrc  # or source ~/.zshrc

Verify the installation (if a version is printed, installation succeeds):

Bash
hermes --version

After installation, the following screen appears:

After installation, the following screen appears:

3. Configure access credentials

Security Reminder

Never share your secret key or embed it directly in browser/client-side code.

If you suspect leakage, delete it and regenerate immediately.

  • Pay-as-you-go/model packages: Go to Console > API Key Management > Create API Key to get the API Key

Pay-as-you-go supports all multimodal models, including text, speech, vision, and more.

  • U Plan: Go to Console > Subscription Management > U Plan to get the API Key

The plan-exclusive Key is used for plan quota. It can exist before paid resources are available; when you have plan quota, that quota is used first. See U Plan Subscription Overview

4. Configure U2-Flash in Hermes Agent

In the two configuration methods below, replace BASE_URL and API_KEY with the actual values you obtained.

Method 1: enter the following commands in online editing for quick configuration

The value of model.provider must be set to custom. Any other custom name is invalid, such as unisound-coding.

Bash
hermes config set model.provider custom
hermes config set model.base_url BASE_URL
hermes config set model.api_key UNISOUND_API_KEY
hermes config set model.default u2-flash

After configuration, you can inspect the settings in ~/.hermes/config.yaml.

Method 2: edit the configuration file manually

Manually edit ~/.hermes/config.yaml:

YAML
model:
  provider: custom
  base_url: BASE_URL
  api_key: UNISOUND_API_KEY
  default: u2-flash

Verify configuration

After configuration, run the following command to send a test message:

Bash
hermes chat -q "Hello"

If a normal reply is returned, the configuration is successful. To switch models, use the -m parameter:

Bash
hermes chat -m qwen3.7-max

5. Model Skill

We provide built-in U2-ASR, U2-TTS, U2-TTS-Clone, and U2-OCR capabilities. You only need to add the corresponding model skill in Hermes Agent to quickly call these model services without integrating each model API separately.

Click the links below to open the corresponding Skill pages. You can install each Skill via chat or the command line. After installation, the Skill will guide you through configuring the related models.

Model
Skill Name
Skill Description
Action
U2-OCR
u1-ocr-parser-pro
Document Parsing
U2-OCR
u1-ocr-extract-pro
Document Information Extraction and Classification
U2-OCR-Med
u1-ocr-med-pro
Medical Document Information Extraction and Classification
U2-ASR
u2-asr-pro
Speech Transcription
U2-TTS
u2-tts-pro
Speech Synthesis
U2-TTS-Clone
u2-tts-clone-pro
Voice Cloning