Guides

How to run a fleet of AI coding agents on hardware you own, without becoming an infrastructure engineer first.

Set up your agent server

Turn a Mac mini into an AI agent server (5 minutes, free)

Install once, then run Claude Code, Codex, and OpenClaw contained on a mini you leave in the closet.

Run AI coding agents on a Linux server

A VPS or a box in the closet, headless. One command, a systemd service, reachable over Tailscale.

Run AI coding agents on your own cloud VM

Self-host in your own AWS, GCP, or Hetzner account instead of renting a sandbox by the second.

Run AI coding agents locally, contained

On the machine you already use, each agent walled off from your files, keys, and other projects.

Control & audit

Give an agent scoped access to one folder

Deny-all by default. Grant a single folder, read-only, through a broker that logs every crossing.

Audit what your coding agent did

Read, verify, and export a tamper-evident ledger of every file an agent touched.

Build on it

Get a phone push when your agent finishes

One command wires Dejima's webhook events to a push on your phone. Add a small server if you want them formatted.

Going deeper

You're still SSHing into tmux to run your agents?

Where the tmux + SSH setup breaks down, and how to run a contained fleet on the same box.

More guides are on the way: running headless agents and building on the API. Want one written? Ask on GitHub. Comparing tools? See how Dejima compares.