A fully open-source agent that learns from every task. It writes its own skills, schedules its own work, and keeps getting better the longer you run it.
By Nous Research · OpenClaw-compatible · WSL2 + Linux + macOS
| Axis | Hermes Agent | OpenClaw |
|---|---|---|
| Cost | Free — MIT license | Free core + $15 guide / subs |
| Best for | Researchers, tinkerers, Linux-natives | Builders who want it running today |
| Learning | Self-improving by default | Memory layers + manual skill install |
| Messaging | Via plugins — some assembly required | 6 channels out of the box |
| Setup | ~20 min, Linux/WSL2 preferred | ~30 min, all platforms |
| Philosophy | Open weights, open everything | Bring your own model, bring your own stack |
Hermes requires WSL2 (Windows Subsystem for Linux). Native Windows is not supported. Set it up with wsl --install before continuing.
Pull the official Nous Research source. The repo includes installation scripts for macOS and Linux.
git clone https://github.com/nousresearch/hermes-agent cd hermes-agent
Installs Python dependencies, downloads the base models, and sets up the local vector store for skills.
./install.sh
Hermes works with any OpenAI-compatible API — OpenRouter, Anthropic, a local Ollama server, or a self-hosted vLLM instance.
cp config.example.yaml config.yaml # then edit config.yaml with your keys
Launches the Hermes loop. It boots, loads its skill library, and starts listening on port 8080.
hermes start
If you already have an OpenClaw setup, Hermes can import your SOUL.md, MEMORY.md, and installed skills automatically.
hermes migrate --from ~/.openclaw
Hermes is MIT-licensed and built by Nous Research. Clone it, fork it, modify the loop — no permission required.