Hermes Agent is an open-source, self-improving AI agent built by Nous Research. It can learn from experience, create skills autonomously, and interact through Telegram, Discord, and CLI.

Hermes Agent itself is lightweight — it runs on any Linux server, macOS, WSL2, or even Android Termux. The AI model runs remotely via API (OpenRouter, OpenAI, etc.), so no GPU required.

Over 200 models including GPT-4, Claude, Llama, Mistral, and any OpenAI-compatible API. You can switch models on-the-fly.

Hermes can spawn isolated sub-agents for parallel tasks. Each sub-agent has its own context and can run independently while the host coordinates scheduling.

Yes. Hermes runs on your own infrastructure. Your conversations and data stay on your server. No data is sent to third parties except the AI model API you choose.