Hermes Agent

Hermes Agent

Hermes Agent: The Ultimate Open-Source AI Harness

Forget the chatbot-in-a-browser. NousResearch/hermes-agent is a full autonomous agent framework that runs where you already work — your terminal, desktop, phone, and even Telegram/Discord/Slack.

What it is

Hermes is in the same league as Claude Code and OpenAI Codex, but provider-agnostic and self-improving. Built by Nous Research, it works with 20+ LLM providers (OpenRouter, Anthropic, OpenAI, Google, xAI, local models via llama.cpp/Ollama) and runs on Linux, macOS, Windows, and WSL.

Same agent, many surfaces:

  • CLI & TUI - terminal-native coding and task execution
  • Hermes Desktop - native Electron app with streaming chat, file browser, and preview pane
  • Gateway - turn the same agent into a bot on 15+ messaging platforms
  • Web Dashboard & API - manage it from the browser or as an OpenAI-compatible proxy

Why it's different

  1. Skills, not prompts. Hermes saves reusable procedures as SKILL.md files that load on-demand. It literally learns from experience.
  2. Persistent Memory. It remembers you across sessions — preferences, projects, and lessons — not just chat history.
  3. Profiles & Cron. Run multiple isolated agents, schedule autonomous jobs (cron), and delegate parallel sub-agents with delegate_task.
  4. Extensible. Themes, desktop plugins, TUI widgets, MCP servers, and webhooks — all hot-reloadable.

If Claude Code is a great coding assistant, Hermes is the self-hosted, always-on teammate. Open source (MIT), actively developed, and designed to be your agent — not a rented chat window.

Check it out: github.com/NousResearch/hermes-agent