What Is Hermes?

An Agent, Not a Chatbot

Most people's experience with AI is a chat window: you open it, ask a question, get an answer, and the conversation ends when you close the tab. Hermes is built around a different idea. It is an open-source AI agent runtime -- software that runs persistent, autonomous agents on your own machine, with their own identity, their own memory, and their own ability to act on a schedule or in response to messages, not just when you happen to be typing.

The distinction matters because it changes what you can build. A chatbot answers the question in front of it. An agent running under Hermes can check Hacker News every morning and text you a summary, watch a server and page you when something breaks, or sit in your team's Slack and answer questions about your codebase -- all without you opening a window to ask it to.

This track (Hermes Agent Fundamentals) gets you from nothing installed to a real, personalized, working agent on your machine. The companion track, Hermes Agent Advanced, picks up from there: scheduling its work, connecting it to messaging platforms, and extending what it can do. You do not need Advanced to get real value out of Fundamentals -- by the end of this track you will have an agent that knows your projects and remembers what matters to you.

The Building Blocks, Briefly

Three concepts come up constantly in Hermes documentation and in this track. You will go deep on each of them later, but it helps to have the shape of them now:

• Profiles. Each agent identity lives in its own isolated directory under ~/.hermes/profiles/<profile>/, with its own configuration, memory, and skills. You can run one agent or a whole fleet of differently-named agents side by side. Covered in Lesson 4.
• The gateway. A background process that connects Hermes to messaging platforms -- Telegram, Discord, Slack, and many more -- so an agent can be reached and can act without you running a terminal session. This is an Advanced-track topic, but it is worth knowing the word now.
• SOUL.md and memory. Hermes agents are not stateless. A SOUL.md file gives an agent a consistent identity and voice, and a memory system lets it retain facts about you and your environment across sessions. Covered in Lessons 5 and 8.

What People Actually Use It For

The Hermes documentation collects real user stories from its community. They cluster into a handful of categories, and they are a useful way to understand what "an agent that persists" is actually good for, beyond the abstract pitch:

• Personal assistance. Summarizing inboxes, managing tasks across tools like Obsidian and Apple Calendar, journaling automation, fitness coaching with nutrition and sleep tracking, medication reminders.
• Information retrieval. Daily Hacker News summaries, research briefs, trending-content curation delivered on a schedule rather than pulled on demand.
• Development. Code review workflows where, as one user put it, the agent "knows the codebase better than the developer after 10 days." Parent agents orchestrating child agents for parallel work. Running models on personal hardware with Ollama or llama.cpp.
• Business operations. Ticket triage and project management automation across tools like Plane.so, lead generation, competitor analysis, CRM lookups via Hunter.io and Google Workspace.
• Content creation. Autonomous social posting, video research workflows, podcast generation, and agents that learn a specific writer's voice and keep producing in it.
• Research and analysis. Market intelligence, prediction-market trading analysis, academic and drug-discovery workflows, ML experiment tracking.
• Home and IoT. Smart home control via Telegram, server and VPS administration, even remote vehicle control and EV battery monitoring.
• Cost-conscious self-hosting. Several stories describe running agents on a $10/month VPS instead of expensive hardware, using smart model routing (cheap models for routine tasks, premium models for hard ones) to keep token costs down. Privacy-focused users run Hermes entirely locally to keep sensitive data off cloud services, including on Raspberry Pi and NixOS systems.

None of this requires you to be a professional developer. It requires comfort with a terminal and the willingness to edit a couple of text files -- which is exactly what the rest of this track walks through.

What This Track Covers

Hermes Agent Fundamentals is nine lessons, building in order toward one real, working agent:

• Install Hermes and have a real first conversation (Lesson 2)
• Get comfortable with the CLI you will use every day (Lesson 3)
• Understand configuration, secrets, and profiles (Lesson 4)
• Give your agent an identity with SOUL.md (Lesson 5)
• Make it aware of your actual projects with AGENTS.md (Lesson 6)
• Pull files, diffs, and web pages directly into a conversation (Lesson 7)
• Teach it to remember things about you and your environment (Lesson 8)
• Put all of it together in one real walkthrough (Lesson 9)

By Lesson 9 you will have an agent that is not a generic chatbot anymore -- it has a name, a voice, knowledge of your specific projects, and a memory of what you have told it. That is "up and running." Everything after that -- scheduling, messaging platforms, voice, browser control, multi-agent coordination, and production hardening -- lives in Hermes Agent Advanced.