Hermes HA

Installation

Hermes Agent runs anywhere a Python environment or container runtime is available. Pick the install method that matches your target environment.

Linux / macOS / WSL2

curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash

Then reload your shell:

source ~/.bashrc   # or: source ~/.zshrc

Windows

If you plan to use Hermes for coding, terminal automation, and project dependencies, WSL2 is usually the smoother Windows path. Install WSL2, open an Ubuntu/WSL terminal, and run the Linux install command.

Manual install with pipx

pipx install hermes-agent

If pipx is not installed yet:

python -m pip install --user pipx
python -m pipx ensurepath

Then reopen your terminal, or reload your shell configuration.

Docker

docker run --rm -it \
  -v ~/.hermes:/root/.hermes \
  -e OPENAI_API_KEY=$OPENAI_API_KEY \
  nousresearch/hermes-agent:latest

Terminal backends

Hermes supports six execution backends, configurable via hermes config:

  • local — run directly on the host.
  • docker — sandbox tool execution.
  • ssh — delegate execution to a remote host.
  • daytona — ephemeral developer environments.
  • singularity — HPC workloads.
  • modal — serverless GPU / hibernation.

Common switches:

hermes config set terminal.backend docker
hermes config set terminal.backend ssh

Use local for the shortest path, docker for command isolation, and ssh when you want heavy work to run on a remote machine.

Configure a model

After installing, run:

hermes model

The wizard guides you through OAuth login, API-key setup, or a custom OpenAI-compatible endpoint. You can also run hermes setup to configure models and tools in one flow.

Verify

hermes --version
hermes doctor

hermes doctor runs a diagnostic check and prints any missing dependencies or misconfigured keys.

If installation succeeds but your shell cannot find hermes, the new PATH probably has not been loaded yet. Close and reopen the terminal, or run source ~/.bashrc / source ~/.zshrc.