Hermes Agent Tutorial

The Complete Guide to Autonomous AI Operations

1. What is Hermes Agent?

Hermes Agent is an autonomous AI assistant built by Nous Research that runs from your terminal. Unlike chat-based AI tools, Hermes can actually do things — execute shell commands, browse the web, manage files, send emails, control smart homes, deploy code, and much more.

Think of it as a junior developer that lives in your terminal and can handle complex multi-step tasks with minimal supervision.

Key Capabilities

  • Execute shell commands and scripts autonomously
  • Browse the web, fill forms, take screenshots
  • Read, write, and edit files across your system
  • Schedule cron jobs for recurring tasks
  • Connect to MCP servers for extended tool access
  • Send messages to Telegram, Discord, Slack, and more
  • Generate images, videos, and audio
  • Manage GitHub repos, PRs, and CI/CD pipelines
  • Delegate tasks to sub-agents for parallel execution
  • Persistent memory across sessions

2. Installation

Quick Install (Recommended)

# One-line installer
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash

UV Pip Install

# Install via uv pip
uv pip install hermes-agent

# Verify installation
hermes version

From Source (Development)

# Clone the repository
git clone https://github.com/NousResearch/hermes-agent.git
cd hermes-agent

# Install in development mode
uv pip install -e .

Tip: Hermes requires Python 3.11+ and an OpenRouter API key. The installer handles most dependencies automatically.

3. Configuration

After installation, run the setup wizard to configure your API keys and preferences:

# Run interactive setup
hermes setup

# Set your model provider
hermes config set model.provider openrouter
hermes config set model.model anthropic/claude-sonnet-4

# Configure terminal working directory
hermes config set terminal.cwd /your/project/path

Config File Location

Configuration lives at ~/.hermes/config.yaml. Key settings include:

  • model.provider — OpenRouter, Anthropic, OpenAI, or custom providers
  • model.model — Which LLM to use (e.g. anthropic/claude-sonnet-4)
  • terminal.cwd — Default working directory for commands
  • gateway — Web gateway settings for remote access
  • tts — Text-to-speech provider configuration

4. The 42 Built-in Tools

Hermes comes with 42 tools organized into categories. Here are the most important ones:

Terminal & File

💻

terminal

Execute shell commands with timeout control and background mode

📄

read_file

Read files with line numbers and pagination

write_file

Write content to files, auto-creates directories

🔎

search_files

Ripgrep-powered file content and name search

🔧

patch

Targeted find-and-replace edits in files

💾

execute_code

Run Python scripts with tool access (up to 50 calls per script)

Web & Browser

🌐

browser_navigate

Navigate to URLs and get page snapshots

🖱

browser_click

Click elements identified by ref ID

browser_type

Type text into form fields

📷

browser_vision

Take screenshots for visual inspection

Automation & Communication

cronjob

Schedule recurring tasks with cron expressions

💬

send_message

Send messages to Telegram, Discord, Slack, etc.

👥

delegate_task

Spawn sub-agents for parallel task execution

🎧

text_to_speech

Convert text to audio via TTS providers

🎨

image_generate

Generate images from text prompts

📝

memory

Persistent memory across sessions

5. Skills System

Skills are reusable procedures that teach Hermes how to handle specific tasks. There are 952+ community skills covering everything from WordPress management to cybersecurity to creative writing.

Using Skills

# List available skills
hermes skills list

# View a specific skill
hermes skills view wordpress-manage

Popular Skill Categories

  • 🏠 Web Development — WordPress, Next.js, React, Tailwind, Shopify
  • 🛡 Cybersecurity — 400+ skills for security analysis, forensics, pen testing
  • 📊 Data Science — Jupyter notebooks, ML training, model evaluation
  • 📡 Marketing — SEO, content strategy, email campaigns, social media
  • 💻 DevOps — Docker, CI/CD, deployment, monitoring
  • 🎨 Creative — Image generation, video, music, ASCII art

Creating Your Own Skills

# Hermes auto-suggests saving successful workflows as skills
# Skills are stored as SKILL.md files with YAML frontmatter
# Location: ~/.hermes/skills/

6. Cron Jobs & Automation

Hermes can schedule tasks to run automatically — either as LLM-driven jobs or as pure script executions:

# Schedule a daily briefing at 9 AM
hermes cron create –schedule “0 9 * * *” –prompt “Generate daily briefing”

# Run a script every 5 minutes (no LLM, just watchdog)
hermes cron create –schedule “*/5 * * * *” –script “health_check.sh” –no-agent

# List all scheduled jobs
hermes cron list

Note: Cron jobs run in fresh sessions with no conversation context. Prompts must be self-contained. Jobs with --no-agent skip the LLM entirely and just run the script.

7. MCP Server Integration

Hermes has a built-in MCP (Model Context Protocol) client that connects to external tool servers:

# Add an MCP server in config.yaml
hermes config set mcp.servers.my-server.command “npx”
hermes config set mcp.servers.my-server.args ‘[“@my-org/mcp-server”]’

This gives Hermes access to any MCP-compatible tools — databases, APIs, custom workflows, and more.

8. Real-World Use Cases

💰

Business Automation

CRM management, email outreach, pipeline reviews, daily journal generation

💻

Web Development

WordPress management, deployments, SEO audits, content publishing

🔎

Research

Web scraping, data analysis, competitor monitoring, academic paper review

🛡

Security

Vulnerability scanning, log analysis, incident response, threat hunting

📡

Content Creation

Blog writing, social media posts, email newsletters, image generation

🏠

DevOps

Docker management, server monitoring, CI/CD, automated deployments

9. Pro Tips & Pitfalls

  • Cost awareness — Use free models (Gemini Flash Lite) for cron jobs, premium models for complex tasks
  • Gateway restart — Don’t restart the gateway from within an agent session; it kills your connection
  • Cron prompts — Make cron prompts self-contained; fresh sessions have no conversation context
  • Skills first — Always check if a skill exists before building from scratch; 952+ are available
  • Memory limits — Keep memory entries compact; they’re injected into every turn
  • execute_code — Use for 3+ tool calls with logic between them; single calls are better as direct tool invocations
  • Sub-agents — Use delegate_task for reasoning-heavy subtasks; keeps parent context clean

Ready to Get Started?

Hermes Agent is open source and free to use. You only pay for the LLM API calls.

View on GitHub →

Read the Docs →

Want a site for your tutorial? Get in touch →