About
The agent that learns as it works.
Hermes One is an autonomous AI agent platform built around a simple idea: your assistant should get better the more you use it — without handing your data or your model choice to anyone else.
Why we built it
Most AI assistants forget you the moment a session ends, lock you into a single provider, and live behind a wall of configuration. Hermes takes the opposite approach. It carries persistent memory across sessions, models you as a user over time, and exposes everything through a native desktop app — no terminal required.
Under the hood, Hermes runs a closed learning loop. After complex tasks it can author its own reusable skills, then improve them during future use. It delegates work to isolated subagents, schedules automations in plain language, and runs across seven terminal backends — from your laptop to a GPU cluster — costing nearly nothing when idle.
What we believe
Open by default
Hermes is free and MIT-licensed. The whole platform is on GitHub — fork it, audit it, or self-host it on a $5 VPS.
No lock-in
Bring any model from any provider, and switch with a single command. Your agent, your memory, your data — never trapped behind one vendor.
Always improving
A true closed learning loop: the agent curates its own memory, writes its own skills after complex tasks, and refines them as it works.
Lives where you do
Telegram, Discord, Slack, WhatsApp, Signal, and CLI from one gateway, with conversation continuity across every platform.
Star history
Hermes is open source — here's how the community has grown over time.
An independent project
Hermes One is an independent project, built and maintained by the Hermes One team. It's powered by the open-source Hermes Agent from Nous Research, but is a separate, third-party product — not affiliated with or endorsed by Nous Research. It's free for anyone to use, study, and extend.