Hermes Agent
The open-source desktop AI agent from Nous Research that remembers you — persistent memory, tool use and real actions across your apps, on a model and machine you fully control.
ChatGPT
Claude
Microsoft CopilotAbout Hermes Agent
Hermes Agent is Nous Research's open-source answer to the closed AI assistant: a desktop agent that doesn't just answer questions but remembers who you are, learns your context over time, and takes real actions across your apps and files. It's built for people who want ChatGPT-level capability without renting it from someone else's cloud.
The headline feature is persistent memory. Where a stock chatbot forgets everything between sessions, Hermes Agent keeps a durable, local memory store — so it recalls your projects, preferences and past work across platforms and picks up exactly where you left off. Point it at the model of your choice, hosted or fully local, and every memory, credential and file stays on hardware you own.
Because it's open source (MIT-licensed) and self-hostable, Hermes Agent is a genuine alternative to ChatGPT, Claude and Copilot for privacy-conscious builders. Install it, connect your tools and your model, and you get a personal AI agent that grows with you — one you can inspect, extend and run entirely offline.
Key features
- Persistent memory — A durable local memory store means the agent remembers your context, projects and preferences across every session and platform.
- Takes real actions — Tool use and cross-platform actions across your apps and files — an agent that does work, not just a chat window.
- Self-hosted & private — Runs on your own machine with hosted or fully local models, so your memory and data never leave your control.
- Bring your own model — Connect an API like OpenAI or Anthropic, or run a local model — swap anytime with no provider lock-in.
How to install Hermes Agent
The honest read
- Persistent, cross-platform memory that most assistants lack
- Fully open source (MIT) and self-hostable
- Takes real actions via tool use — a true agent
- Works with hosted or fully local models
- Backed by Nous Research and an active community
- Setting up a self-hosted agent takes more effort than opening ChatGPT
- Best experience needs a capable model and some hardware
- Agent actions need guardrails and human review
- Younger project than the biggest closed assistants
FAQ
Hermes Agent is an open-source desktop AI agent from Nous Research. Unlike a normal chatbot, it has persistent memory, uses tools and takes actions across your apps — and you self-host it with the model of your choice, so your data stays on your own machine.
Yes. Hermes Agent is MIT-licensed and free to self-host. Compared to ChatGPT it adds persistent memory and real tool use, and it runs on infrastructure you control rather than a closed cloud.
Hermes Agent keeps a durable local memory store rather than forgetting between sessions. It remembers your projects, preferences and past interactions across platforms, so it stays useful over weeks and months instead of starting fresh each chat.
Yes. You can connect a hosted API or run a fully local model, keeping the entire agent — memory, credentials and actions — offline and private for sensitive work.
Self-host Hermes Agent from the NousResearch/hermes-agent repository, connect your chosen model and tools, and the agent runs on your machine. The setup walkthrough below shows the full process step by step.
Because it takes real actions, review its permissions and keep a human in the loop for sensitive operations. Self-hosting means the agent and your credentials stay on infrastructure you control.
Python
PyTorch
SQLite
DockerSimilar open-source projects
Open-source personal AI assistant that runs on any OS and takes real actions.
Self-hosted chat interface for any LLM, local or API.
All-in-one AI workspace — chat with your documents, any model.