Skip to main content
A personal assistant needs a computer of its own, and a boxd machine is built to be one. Run Hermes, OpenClaw, or any assistant that lives on a box. The assistant’s files, notes, cron jobs, tool logins, and half-finished work all sit on a persistent disk, and it answers webhooks at its own public HTTPS URL.

/boxd-setup-hermes installs the Hermes assistant right on the machine.

Put an assistant on a machine

Any assistant that runs on Linux can live here. Hermes by Nous Research is one example, a self-improving CLI agent with a built-in messaging gateway for Telegram, Discord, Slack, WhatsApp, and Signal, and it happens to be the one boxd ships a skill for. One command installs it on a machine:
Run it from Claude Code on your laptop (it creates or reuses a VM and installs over boxd machine exec) or inside any machine, where the skill comes pre-baked. It runs Hermes’ official installer, then hands you the interactive parts:
The skill can also expose Hermes’ messaging gateway on the machine’s public URL, so your chat apps reach it directly. See Skills for what it installs. Claude Code, Codex, and OpenCode also ship on every machine, so ssh mybot.boxd and claude is a personal assistant with a computer too. See Coding agents.
Hermes is an independent project by Nous Research. The skill automates installing it on a boxd machine.

Give it your tools

Connect integrations once at the account level and every personal machine can act as them through the run CLI and the @boxd/run SDK:
  • Slack: run slack send-message --channel "general" --markdown-text "…" to post, slack.on("…") to react to channel events.
  • Linear: run linear list-issues, run linear create-issue, and issue events as triggers.
  • GitHub: once GitHub is connected, or your org has installed the GitHub App, git and gh work against your private repos with nothing else to configure, and run github … covers the rest.
The coding agents on the machine discover these on their own (run lists what’s connected). Anything the assistant should do on a schedule or in reaction to an event belongs in an automation: a *.run.ts file that boxd keeps running, wakes the machine for, and shows on the console. Everything else the assistant needs, like API keys for its LLM provider or your services, goes in once through env vars and secrets and is injected into every machine you own. Secrets are sealed at rest and never readable back.

Keep it awake

If your assistant runs background tasks like scheduled jobs, we recommend keeping its machine always on. Turn hibernation off with: