🪽Hermes Agent
Nous Research's self-improving agent — runs anywhere from a $5 VPS to Telegram, Slack, or WhatsApp, and delegates coding work to Claude Code.
What it does
Hermes is "the agent that grows with you": instead of arriving with a fixed toolset, it creates skills from its own experience, improves them during use, searches its past conversations, and builds a model of who you are across sessions.
What it does:
- Runs everywhere — a $5 VPS, Docker, your laptop, or GPU clusters; talk to it from the CLI, Telegram, Discord, Slack, WhatsApp, or Signal.
- Works unattended — a built-in cron scheduler runs automations while you sleep.
- Parallelizes — spawns isolated subagents for parallel workstreams (one scraping, one summarizing, one cross-checking).
- Any model — 300+ models via OpenAI, OpenRouter, or Nous Portal; switch with
hermes model, no lock-in.
How it pairs with Claude Code: Hermes ships a bundled skill that delegates coding tasks to Claude Code — you ask Hermes for a feature or a PR from your phone, and it drives Claude Code in your terminal to build it. Same SKILL.md format as the skills on this site.
Where does it install? On whichever computer runs the install command below. Your Mac or Linux laptop is the easiest start: open the Terminal app, paste the command from the Installation steps, press Enter. If you want Hermes awake 24/7 — answering you on Telegram while your laptop sleeps — run the same command on a cheap always-on server (a $5/month VPS) over SSH instead. Either way it installs the hermes command on that machine; type hermes to start it.
Connect it to Claude Code (optional, same machine): install Claude Code with npm install -g @anthropic-ai/claude-code, run claude once to log in, and you're done — no config. Hermes's built-in Claude Code skill finds the claude command automatically. Then just ask Hermes: "use Claude Code to add error handling to my project" and it drives Claude Code with turn and budget caps.
Installation
- Open the Terminal app on the machine it should live on.
Your Mac or Linux laptop is fine to start (Terminal lives in Applications → Utilities on a Mac). If you want it running 24/7, SSH into an always-on server and run the same steps there instead.
- Paste into your terminal.
Press Enter and follow any prompts from the installer. It sets up the agent's own command-line tool on this machine.
- Start it.
Run the command the install output shows (usually the agent's own name) in the same terminal. No Claude Code restart needed — this agent runs outside Claude Code.
- Use it.
Talk to it in its own interface — the terminal UI, or whatever channels it supports (see the description above). Check the source README for its commands.
Prefer to read the source first? Open on GitHub.