Master Hermes Agent: From Installation to Advanced AI Automation

Stage 1: Getting Started

The goal is to install Hermes and complete the first dialogue.

Installation

macOS, Linux, and WSL2 can be set up with a single command:

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

Windows users must first install WSL2 and then run the same command inside the WSL2 terminal.

The installer automatically handles Python, Node.js, and Git dependencies and creates the configuration directory ~/.hermes/. Verify the installation with:

hermes version
hermes doctor

Model Configuration

Run the configuration wizard and select "Quick Setup". Beginners can start with free models from DeepSeek, Kimi, or OpenRouter. hermes model Test the model with a simple query:

hermes chat -q "Hello! What tools do you have available?"

Common Commands

hermes

– start a conversation hermes model – select or change the model hermes tools – configure tools hermes gateway – start the message gateway hermes setup – re‑run the configuration wizard hermes doctor – diagnose problems hermes migrate openclaw – migrate from OpenClaw

After these steps Hermes should be fully operational.

Stage 2: Advanced Configuration

Turn Hermes from "usable" into "highly usable" by configuring modules and memory.

Modular Model Setup

Hermes splits tasks into eight modules; each module can use a different model. Use premium models (e.g., Claude Opus or GPT‑4o) for high‑intelligence modules such as memory summarization and skill creation, and free OpenRouter models for routine text processing to reduce token costs.

SOUL.md – Personality File

The file ~/.hermes/SOUL.md defines the agent’s identity, speaking style, preferences, and problem‑solving approach. Two ways to create it:

Manual edit: Directly edit ~/.hermes/SOUL.md and write specifications (e.g., "prefers data‑first decisions", "communicates in Chinese", "backend developer").

Self‑learning: Converse with Hermes, then ask it to "adjust SOUL.md based on our dialogue"; the agent will summarize your preferences automatically.

Review the result with:

cat ~/.hermes/SOUL.md

Memory System

The memory subsystem consists of five files: SOUL.md – personality definition (user‑maintained). user.md – long‑term observations automatically written by Hermes. memory.md – event memory injected into each conversation. skills/ – reusable workflow definitions that grow over time. state.db – low‑level storage of chat logs.

Only SOUL.md requires manual attention; the other files are managed by Hermes.

Platform Integration

Hermes supports more than 15 messaging platforms. Example integrations:

Feishu (Lark): hermes gateway setup Provide the App ID and App Secret obtained from the Feishu Open Platform, then start the gateway: hermes gateway Sending a message to the bot in Feishu confirms successful connection.

Telegram: Create a bot with @BotFather, obtain the token, then configure it:

hermes config set TELEGRAM_BOT_TOKEN your-bot-token

Enable the adapter in the configuration (optional YAML snippet):

gateway:
  adapters:
    telegram:
      enabled: true

Start the gateway: hermes gateway A reply from the bot verifies the integration.

Stage 3: Advanced Usage

Enable Hermes to evolve autonomously and run automated tasks.

Skill System

After completing a complex task, Hermes can package the experience into a reusable Skill. Subsequent similar tasks invoke the stored Skill, improving efficiency.

Trigger Skill creation by completing multi‑step workflows, explicitly requesting "create skill", or repeating a pattern across sessions. ls ~/.hermes/skills/ Example: ask Hermes to fetch AI news from Hacker News and summarize it in Chinese.

hermes chat
> Please fetch the AI news from Hacker News homepage and provide a Chinese summary.

Cron Tasks

Hermes includes a built‑in Cron scheduler. Edit ~/.hermes/cron/tasks.yaml to define recurring jobs.

tasks:
- name: daily_news
  schedule: "0 9 * * *"
  command: "Summarize today’s AI news"
  platform: telegram
  chat_id: "your-chat-id"

The task runs every morning at 09:00 and posts the result to the configured platform.

Troubleshooting & Migration

Run hermes doctor to locate configuration errors, model connectivity issues, or permission problems.

Migrate from OpenClaw with a single command that transfers skills, memory, and configuration: hermes migrate openclaw Confirm migration success by restarting Hermes and asking "Do you know who I am?" – a correct answer indicates that the memory and personality have been imported.