Sparround

Installation routes and what gets created

There are four official routes:

  • Desktop installer (macOS/Windows) — visit the site, download the file, run it
  • Shell installer (Linux/macOS/WSL2/Termux) — a one-line curl ... | bash
  • PowerShell installer (Windows) — iex (irm ...)
  • Docker — a prebuilt image

Apart from Git, the installer brings everything itself: uv, Python 3.11, Node.js, ripgrep, ffmpeg, the virtual environment and the global hermes command.

A one-line curl | bash is convenient, but it means downloading a script and executing it immediately. In a corporate environment the normal practice is: download the script to a file, read it, then run it. We come back to this later — Hermes itself flags the curl | sh pattern as a dangerous command.

WhatPer-user installRoot install
Code`~/.hermes/hermes-agent/``/usr/local/lib/hermes-agent/`
Binary`~/.local/bin/hermes``/usr/local/bin/hermes`
Data and config`~/.hermes/``~/.hermes/`

The ~/.hermes/ directory is the agent's whole life — knowing it makes everything later easier:

  • config.yaml — every non-secret setting
  • .env — API keys, tokens, passwords
  • auth.json — OAuth credentials
  • MEMORY.md, USER.md, SOUL.md — memory and identity
  • skills/ — installed and agent-authored skills
  • sessions/, state.db — conversation history
  • logs/ — logs (secrets are redacted automatically)
  • cron/ — scheduled jobs and their output

Practice. After installing, run hermes doctor and look through ~/.hermes/ with ls -la. Done means: you have checked the permissions on .env (chmod 600 is recommended) and can explain what each file is responsible for.

📚 Sources and documentation