The Complete Guide to Hermes Agent
Deploy Hermes once on your server. It learns your projects, builds its own skills, and stays reachable on every device you use — quietly getting smarter every day.
Just a chat. Lots of jobs.
One persistent agent on your server — long-term memory, real browser control, scheduled jobs, and a self-evolving Skills system.
Tell It Once, It Runs Forever
Describe a recurring job once. Hermes runs it on schedule, survives restarts.
Knows How You Work
Teach it your conventions once. Every future task ships in your voice.
Actually Opens The Page
Hermes drives a real browser — the page itself, not a search snippet.
Remembers Months Back
Stop re-explaining yourself. Hermes pulls up decisions and reasoning from chats months back.
Get Your 24/7 AI Assistant in 5 Minutes
Install, configure, and start using Hermes in one short pass.
Migrating from OpenClaw?
Done in 3 steps
Use hermes claw migrate for one-shot migration — configs are preserved. Backup, import, diagnose in one flow.
Not a Copilot. Not a Chatbot.
Hermes Agent is an open-source autonomous AI agent from Nous Research. Not an IDE copilot or a thin chatbot wrapper, but a long-running agent that lives on your server, keeps what it learns, and becomes more capable over time.
Frequently Asked Questions
Quick answers to the questions we get most often.
HermesAgent 101 is an independent field guide to Hermes Agent — the open-source autonomous AI agent from Nous Research. We focus on practical, real-world workflows: get a working setup in 5 minutes, then go deep on skills, memory, MCP, gateways, and deployment.
Yes. Hermes itself is open source under the MIT license — free forever. You only pay for the language model you point it at (OpenRouter, Anthropic, OpenAI, DeepSeek, etc.), and you can switch providers in one command. Local models via Ollama or vLLM are free end-to-end.
Hermes runs as a long-lived daemon on your own server, with built-in memory, skill evolution, and gateways for Telegram, Discord, Slack, and email. Claude Code is a terminal IDE assistant tied to Anthropic models. OpenClaw is an earlier fork that Hermes has largely succeeded — see our full comparison and one-shot migration guide.
Native Windows support is experimental. For a stable setup, install WSL2 first and run Hermes inside it — the install script handles everything from there. Linux and macOS are first-class.
No. Hermes runs entirely on your own machine. The only data leaving your server is what you send to the LLM provider you choose — and you can route to a fully local model (Ollama, vLLM, SGLang) if you don't want any external calls at all.
About 5 minutes for a working chat on Linux or macOS. Add another 5 if you want a Telegram or Discord gateway. Going end-to-end through the learning path — skills, MCP, cron, deployment — usually takes one evening.
Yes. We welcome corrections, new tutorials, suggested skills, and bug reports. Email us at .
Stay Current on Hermes
Only the updates that are actually worth your time.