Deployment Options

Deploy OpenClaw using various methods

OpenClaw can run on your own hardware, a VPS, managed hosting, in Docker, or on serverless runtimes. The right choice depends on uptime, how much you want to operate, and whether you need the full tool stack (browser, exec, cron). Start with the decision guide below—then open the linked walkthrough.

Where to run OpenClaw

Use this table to pick a host shape. All paths assume you will harden DMs and the Gateway (hardening, pairing).

Option Best when Tradeoffs Guide
Local laptop / desktopTrying OpenClaw, dev, file-heavy workflows on your daily machineSleeps; not 24/7 unless you keep it awakeInstall Ā· No-terminal
Mac Mini home server24/7 at home, Apple Silicon, you can leave hardware onHome IP, power, sleep updatesMac Mini guide
VPSPublic uptime, teams bot, predictable remote hostYou patch, monitor, pay monthlyVPS hub Ā· Compare providers
Managed hostingChannels live fast, minimal opsLess control, vendor pricingManaged guide
BYOSYour VPS account, someone else installs OpenClawStill your infra billBYOS
Docker on a serverReproducible deploy on VPS or homelabContainer ops on top of host opsDocker

Gateway at home, you elsewhere? Use Remote access (Tailscale, SSH tunnels). Serverless subset? See Moltworker—no browser/exec. Cost ranges: How much does OpenClaw cost? Ā· After install: Personal assistant checklist.

Running production-style upgrades? Use the dedicated Deployment Upgrade Runbook for preflight checks, smoke tests, and rollback flow.

Want a hands-on tutorial? Follow our step-by-step Deployment Tutorial to learn how to deploy OpenClaw with Docker. This page provides an overview of deployment options.

šŸ’¼ Workplace channels

Slack, Microsoft Teams, Discord—quickstarts for Socket Mode, Azure HTTPS, and pairing before you open the firewall.

Choose a channel →

šŸ–„ļø VPS Deployment

Run OpenClaw on a virtual server—Hetzner, DigitalOcean, Vultr, Hostinger. Full control, 24/7 uptime. Compare providers and follow step-by-step setup guides.

VPS Guide →

šŸŽ Mac Mini home server

Run OpenClaw 24/7 on Apple Silicon at home—sleep settings, remote access, and when a VPS is simpler.

Mac Mini guide →

🐳 Docker Deployment

Deploy OpenClaw using Docker containers. Learn containerization, orchestration, and best practices for Docker-based deployments.

Docker Guide →

ā˜ļø Serverless Deployment

Deploy OpenClaw on serverless platforms like Cloudflare Workers using Moltworker. Learn about serverless architecture and limitations.

Moltworker Guide →

šŸ  Managed Cloud

Run OpenClaw without managing servers. Clawi, xCloud, MyClaw.ai, OpenClaw Cloud—WhatsApp, Telegram, Discord in minutes. No VPS, no SSH.

Managed Hosting Guide →

šŸ“¦ BYOS (Bring Your Own Server)

You provide the VPS; a provider installs and manages OpenClaw. OpenClawHosting.io, MoltBotHost, Elest.io, ClawRun.dev. Data on your chosen infra.

BYOS Guide →

šŸ¢ Enterprise & Teams

Multi-agent teams, shared workspaces, and team deployment. SSO, audit logs, and scaling patterns for organizations.

Enterprise →

Gateway logs and observability

Most installs are a long-lived Gateway plus channels. Observability means you know when the process restarted, why a message failed, and whether disk or CPU is trending wrong—without leaking secrets into logs.

Where logs come from

On a VPS you might use systemd, pm2, or Docker. The runtime captures stdout/stderr. Identify which unit or container owns the Gateway, then tail that stream only.

Redaction and retention

Assume logs will be pasted into a ticket: mask tokens, API keys, webhook secrets, and phone numbers before sharing. Chatty channels and debug flags fill disks—set rotation (journald limits, Docker log options, or logrotate). Pair logging with security hardening so you are not widening the attack surface for convenience.

Health checks and export

At minimum watch process up and the Gateway TCP port. For richer signals, reuse incident alerting. Shipping lines to Grafana Loki, Datadog, or syslog is normal plumbing—point your collector after redaction rules exist. Flags and formats drift—check docs.openclaw.ai on upgrade. See also Remote access and Deployment upgrade runbook.

Enterprise & Team Deployment

For teams and organizations, OpenClaw can run as multiple agents with shared context, per-channel sandboxing, and controlled DM access. Key patterns:

  • Multi-agent squad — Separate agents for different roles (support, research, ops). Multi-agent team tutorial
  • Shared context — One workspace or symlinked directory so agents read from the same page. Shared context use case
  • Sandboxing — Set agents.defaults.sandbox.mode to "non-main" so group sessions run in isolated containers
  • DM policy — Keep pairing for unknown DMs; use allowFrom to restrict who can message

For managed enterprise options, see Managed Hosting (some providers offer team plans). For self-hosted, use VPS or Docker with multiple Gateway instances or session routing.

When moving to a new server or reinstalling, back up config, credentials, and workspace—see Workspace Backup & Restore.