Table Of Contents

Versions

Installation

Initial Setup

First chat with the Agent

Congrats, You made it work !

Extra : Things to consider OpenClaw Challenge Submission 🦞 This is a submission for the OpenClaw Writing Challenge In this article, I want to guide you on how to make OpenClaw "just work" on Windows. I used Windows as my OS, but I think this guide can be easily followed on Linux as well. Chances are, I don't have to explain the title. If you (like me) are trying to participate in the OpenClaw challenge or just trying OpenClaw for the first time, you certainly found that making OpenClaw "just work" on your machine is not an easy task. I think there is room to improve in this area for the OpenClaw team, or maybe that's the point of this challenge after all? hehe You tell me. After that fresh introduction, I will try to be direct and simple in this guide. I think I need to save you some time since you probably already wasted some, trying to set up OpenClaw... For reference and reproducibility: Note: I use a specific version for reproducibility. The team is making releases every day and something might get broken if you try a different version after the publication date of this post. I did experience issues with version 2026.4.21, so if you are skilled enough, feel free to play with the latest version; but if you are in the mode of "let's make this thing just work," stick to the guide. To set up OpenClaw, we need to set a model and model provider; this is necessary for the agent to work. As the provider, we will use Google AI Studio. Let's go to Google AI Studio and look for the section to generate an API KEY. It is relatively easy to do. Once we have it, we save it in a secure place. It's all we need beforehand; we can proceed with OpenClaw. Let's start with the command: Now, on my computer, I had to wait about 4 minutes for the command to show the CLI of OpenClaw. Haha, this is just a warning in case you run the command and see it hangs. Just trust the process and wait for it. It seems it needs to install something else before actually starting. After some minutes you should see something like: So, you are about to set up. Here are the default answers to use to finish the setup quickly: Ok, the setup is ready. Let's confirm by looking at the default model, run: You should see something like: It is important that the default model is the one we set during onboard. If everything is as expected, we can proceed to play with the agent. In one terminal run : This is like "run the server" to start playing with an agent. In another terminal run: This is like "open a client" to chat with the agent. Once is ready type something, like "Hello". If you are lucky enough, you should receive a response: And there you go. You can start playing and interacting with OpenClaw. So now you can chat with the agent. The agent will guide you on how to define the behavior, the personality, and the tasks to do. You can continue exploring on your own from this point. Hopefully, this tutorial was useful and saved you some time :) Templates let you quickly answer FAQs or store snippets for re-use. Hide child comments as well For further actions, you may consider blocking this person and/or reporting abuse