Installing OpenClaw can be a fast process if you choose the right path from the start. The official installation flow is designed to take you from the terminal to a working chat in just a few minutes, as long as you are clear on the operating system, model provider, and Gateway configuration.
In this guide, we will cover how to install OpenClaw using the official methods, what happens during onboarding, how to validate that the service is active, and when it makes sense to go straight to a prepared OpenClaw Hosting setup. The goal is not to copy commands without context, but to give you a safe path for labs, testing, or production.
What OpenClaw is and why install it on your own infrastructure
OpenClaw is an open source AI assistant that you can run on your own infrastructure. That makes it attractive for technical teams, founders, and advanced users who want control over the environment where their agent runs, the model provider it uses, and the services connected to it.
The key decision is where to run it. You can install OpenClaw on your local machine for learning and debugging, or deploy it on a VPS if you need the agent to stay continuously available. Teramont’s own OpenClaw hosting proposal presents it as an AI assistant running 24/7 on infrastructure dedicated to the service: https://teramont.net/es/openclaw-hosting.
If OpenClaw needs to respond, keep a Gateway active, or run tasks without depending on your laptop, an always-on server is no longer optional.
Prerequisites before installing OpenClaw
Before running the installer, it is worth preparing three things: operating system, runtime, and AI provider credentials. The getting started documentation indicates that OpenClaw aims to take you from installation to a functional chat in around five minutes, with Gateway and authentication configured: https://documentation.openclaw.ai/start/getting-started.
Runtime and operating system
OpenClaw uses Node. According to the getting started guide, Node 24 is the recommended version, while Node 22.19+ remains supported: https://documentation.openclaw.ai/start/getting-started. The official installer can detect the environment and install Node if needed, which reduces common errors on new systems: https://docs.openclaw.ai/install/installer.
The CLI path is documented for macOS, Linux, native Windows, and WSL2, so you can use it both on a Linux VPS and on a Windows workstation prepared for development: https://docs.openclaw.ai/start/onboarding-overview.
Model provider credentials
During onboarding, you will need to configure the model provider and its authentication. In practice, that means having the API key or equivalent mechanism available before starting the setup assistant. If you do not have it, you can install the CLI, but you will not be able to correctly complete the agent’s functional flow.
Fast official method: install.sh or install.ps1
The recommended path to get started is to use the official installer. On macOS and Linux, you use install.sh; on Windows, you use PowerShell with install.ps1. The installer documentation explains that it detects the operating system, prepares dependencies when appropriate, and launches the onboarding flow: https://docs.openclaw.ai/install/installer.
On a Linux VPS, the general sequence is to download the installer from the official guide, give it execution permissions, and run it:
chmod +x install.sh
./install.shOn Windows, execution is done from PowerShell with the corresponding script:
powershell
.\install.ps1Once it finishes, continue with the full onboarding. The key command is:
openclaw onboard --install-daemonThis command configures the model provider, authentication, Gateway, and background service in a single walkthrough. The onboarding documentation indicates that this flow usually takes around two minutes: https://docs.openclaw.ai/start/onboarding-overview.
Alternative installation for advanced users
If you do not want to use the automatic installer, the official documentation also covers installation through package managers and from source. According to the repository’s installation index, OpenClaw can be installed with npm, pnpm, and bun, as well as built from a Git clone: https://github.com/openclaw/openclaw/blob/main/docs/install/index.md.
| Method | When to use it | Main advantage |
|---|---|---|
| install.sh / install.ps1 | First installation or clean VPS | Detects the environment and simplifies dependencies |
| npm, pnpm, or bun | Already controlled Node environments | More familiar for JavaScript teams |
| Git clone and local build | Development, testing, or code review | More control over the installed version |
| install-cli.sh | Isolated installation under a local prefix | Keeps Node and OpenClaw inside a directory such as ~/.openclaw |
The practical difference between install.sh and install-cli.sh is isolation. The general installer is intended to make startup easier on the current system. By contrast, install-cli.sh keeps OpenClaw and Node under a local prefix, which is useful if you do not want to touch a global Node installation or if you manage multiple environments on the same machine: https://github.com/openclaw/openclaw/blob/main/docs/install/index.md.
What OpenClaw onboarding actually does
Onboarding is not a decorative form: it is the part that turns an installed CLI into a usable agent. The documentation states that it configures the model provider and its authentication, the workspace, the Gateway, optional channels, and the daemon: https://docs.openclaw.ai/start/onboarding-overview.
In practical terms, during this step you decide which AI provider OpenClaw will communicate with, where it will work, how it will expose or keep the Gateway active, and whether it will run as a background service. If your use case requires messaging channels, such as a flow connected to Telegram or WhatsApp, treat them as optional extensions that you should activate only if they appear available and supported in your installation.
For a server, the most important flag is --install-daemon. Without the daemon, you can run OpenClaw interactively, but you will depend more on the active session. With the daemon, the process is prepared to work as a service, which makes more sense on a VPS or dedicated server.
Verification after installation
Do not consider the installation complete until you run the basic checks. The documentation recommends validating the version, general status, and Gateway with these commands: https://github.com/openclaw/openclaw/blob/main/docs/install/index.md.
openclaw --version
openclaw doctor
openclaw gateway statusopenclaw --version confirms that the CLI is available. openclaw doctor reviews the configuration and helps detect environment issues. openclaw gateway status tells you whether the Gateway is active, which is a critical check if you expect OpenClaw to receive or process requests.
Common issues
The
openclawcommand does not appear: check Node, the global prefix, and the PATH, as suggested by the official guide.Onboarding does not finish: confirm that the model provider API key is valid and that authentication was completed.
The Gateway does not respond: run
openclaw gateway statusand thenopenclaw doctorto isolate whether the issue is related to the service, configuration, or environment.Windows creates friction: try native Windows if your environment is ready; if you manage Linux tools or server dependencies, WSL2 may be a more convenient path.
OpenClaw on a VPS: when it makes sense
Installing OpenClaw locally is useful for learning, reviewing the CLI, or testing a model provider. But if you want continuous availability, less dependence on your personal machine, and an environment closer to production, a VPS for OpenClaw is usually the logical option.
On a VPS, you can control resources, networking, location, and persistence. You also reduce the risk of the agent stopping because you closed your laptop, your local network changed, or you restarted a development session. If you also need to evaluate CPU, RAM, NVMe storage, or connectivity, you can review Teramont’s Hardware page to better understand the available infrastructure.
Fast alternative: preconfigured VPS for OpenClaw
If your priority is to get OpenClaw running without manually installing dependencies or logging in through SSH just to start the service, Teramont offers a specific OpenClaw Hosting option. The page indicates automatic configuration, deployment in under 60 seconds, NVMe storage, 1 Gbps network, DDoS protection, and 24/7 support: https://teramont.net/es/openclaw-hosting.
This path fits when you need to validate an idea, deliver an environment to a team, or prevent manual installation from becoming the bottleneck. Instead of preparing Node, checking PATH, running installers, and configuring the service, you start from an instance designed for OpenClaw from the beginning.
Also, if you already have a local instance or one on another server, the same page indicates that you can migrate it to the VPS and get assistance from the team to move configuration, memory, and skills: https://teramont.net/es/openclaw-hosting.
Which path to choose
Choose manual installation if you want to learn each component, debug errors, customize the installed version, or work from source. It is the best option for understanding how the CLI, Gateway, daemon, and model provider connect.
Choose a preconfigured VPS if the goal is to have OpenClaw available quickly, with fewer operational steps and a more production-ready foundation. As a final checklist: install or deploy the instance, complete onboarding, validate with openclaw doctor, check openclaw gateway status, connect only the channels you need, and scale infrastructure when usage justifies it.
