Hermes Agent is Nous Research’s self-hosted personal agent, released in February 2026 under the MIT license. It runs on hardware you control rather than in a vendor’s cloud, and it is built around a learning loop: Hermes keeps small, bounded memory files about you and your work, and after it repeats a task it writes its own reusable skills, so the agent measurably gets better at your workflows the longer it runs.
Hermes is model-neutral: connect Nous Portal, OpenRouter, OpenAI, Anthropic, Google or any OpenAI-compatible endpoint, and swap models per agent. The same agent, with the same memory and tools, reaches you through more than twenty messaging platforms, including Telegram, Discord, Slack, WhatsApp, Signal and email, as well as a terminal app and a desktop app for macOS, Windows and Linux. It can schedule jobs in natural language, spawn isolated subagents for bigger tasks, search and browse the web, and run in sandboxed backends from local and Docker to SSH and Modal. Your conversation state and data stay in a folder on your own machine.
Hermes Agent is free to self-host under the MIT license; you pay only for your own model calls. If you want the hosted route, Nous Portal offers a free tier plus paid plans with monthly credits and access to 200+ models: Plus at $20 per month, Super at $100 and Ultra at $200.
Hermes Agent fits builders and privacy-conscious power users who want a long-running personal agent that learns how they work, without handing their data or their model choice to a single vendor.