Run Kiro Crew 24/7

Kiro Crew is a persistent agent that runs on your own hardware rather than in a session you open and close. Running it continuously means installing it as a service, or as a container, and reaching its dashboard through a tunnel.

When to use this

Run Crew 24/7 when you want scheduled jobs, reactive monitoring, and background subagents that keep working while you do something else. A laptop that sleeps is a poor host. A small always-on Linux box, or a cloud instance you already run, is a good one.

Steps

1. Check the prerequisites. Crew needs Python 3.10 or later, with 3.12 recommended. Node.js 22 or later is only needed to build the dashboard from source, so the wheel, desktop, and Docker paths do not need it. Model access goes through kiro-cli, which the desktop app installs on first launch.

2. Install it. Pick one:

   curl -fsSL https://download.crew.kiro.dev/cli.sh | sh

   pip install kirocrew

   The desktop app is also available as a signed .dmg for macOS, a signed Windows installer, and .deb, .rpm, or AppImage packages for Linux, on x86_64 and ARM64.

3. Run the wizard and verify:

   kirocrew setup
   kirocrew doctor

4. Start the gateway once by hand to confirm it works. kirocrew gateway serves the dashboard at http://localhost:5476. Data lives in ~/.kiro/crew/, which KIROCREW_HOME can relocate, and the port is set by KIROCREW_PORT.

5. Install it as a service so it survives reboots:

   kirocrew service install
   kirocrew service status
   kirocrew logs -f

   This registers systemd on Linux and launchd on macOS.

6. Or run it in Docker instead:

   docker run -d --name kirocrew -p 127.0.0.1:5476:5476 \
     -v kirocrew-home:/home/kirocrew \
     ghcr.io/kirodotdev/kirocrew:stable
   docker exec -it kirocrew kiro-cli login

   Note the 127.0.0.1: prefix on the port mapping. It keeps the dashboard off the network.

7. Size the host. A remote machine should run Ubuntu 22.04 or later, Debian 12 or later, Fedora, or Amazon Linux 2023, with about 10 GB of RAM minimum and 16 GB recommended.

8. Reach a remote dashboard through an SSH tunnel, because it binds to loopback only by default:

   ssh -L 5476:localhost:5476 user@host

Check it worked

Reboot the host. Then run kirocrew service status and open the dashboard through your tunnel. Both should work with no manual start. Issue a short-lived access token with kirocrew token --ttl 2h when you need one, rather than leaving a long-lived credential in place.

Common problems

• The dashboard is unreachable. It binds to loopback. Use the SSH tunnel rather than opening port 5476 to the network.
• The service dies after a reboot. Use kirocrew service install instead of starting kirocrew gateway from a shell.
• The port is taken. Set KIROCREW_PORT to something free.
• A second instance behaves oddly. You copied the secret files. Regenerate them with kirocrew setup on that host.
• Third-party apps do not load. They are disabled by default. Enable them under Settings, Security, and only for apps you have read.
• Sandbox mode hides your credentials. That is deliberate. The Auto mode hides .gnupg, .gcloud, .azure, and .docker; Strict also hides .aws, .ssh, and .kube; Off is available on Linux and macOS only. Choose the mode intentionally rather than turning it off out of frustration.

Related

• Kiro Crew explains what Crew is and what it can do.
• Set up permissions.yaml covers the wider permission model.
• Start a cloud session from the CLI is the managed alternative to running your own host.