Introduction: A Free OpenClaw Setup
Here's my OpenClaw setup. It doesn't cost me anything because it's running an open source large language model that's on my own PC. In this video, I'm going to show you how to set that up. This will allow you to run OpenClaw autonomously as a long as you want without an additional bill because you're not going to be using tokens from things like Claude or chat GPT since you're running your own model. Let's begin.
Step 1: Installing Ollama
Our first step is to install Ollama, a local AI model that will run on our own PC. Let's do this now.
Head to Google and search up the word Ollama. This is the software that will run our open source a large language models locally on our own PC. It works on any operating system and you can also install a desktop app, but I'm going to do it the traditional way, which is through a terminal. It's only about 100 megabytes and should download in next to no time. It's now installed. Let's open it.
Head to your applications or just search up Ollama from the quick search. It should open up looking a little bit like this. We don't have any models yet to use, so we're going to select one to download. There's quite a few different options here, but this isn't the ideal way to search or download models. I actually prefer to use the website, so head to models over here.
Step 2: Downloading and Running a Local Model
From here you can filter models by newest, most popular, as well as things like thinking, tools, and vision. My favorite right now is Qwen 3.5 since it's a versatile model that seems to work all round.
So, in order to get this running on Ollama, all we have to do in terminal is run ollama run qwen:3.5. Let's paste this in. This will go ahead and download that large language model and then start allowing you to prompt with it through the terminal. You can also prompt through it directly through the app. So, just open up the Ollama app and then head over here. Make sure that you have Qwen 3.5 selected and you can chat with it kind of like you would with chat GPT.
Step 3: Understanding Model Sizes and Memory
Be aware these open source models can come in different sizes from 1 billion to 120 billion parameters, but it doesn't mean that you'll be able to run them all on your PC. It's worth knowing exactly how much memory your computer can handle and this way you'll know which models you can run. Right now, I'm using a Mac with 64 gigs of physical memory and I'm using the baseline version of Qwen 3.5, which is 9 billion parameters or 6.6 gigs of physical memory. Based on my memory capacity, I can actually move up to the 35 billion parameter model, which uses 24 gigs of memory. It's going to take a little bit more time to download. However, it's going to produce a much better outputs whenever I prompt it.
I wish I could go up to the 122 billion parameter model, but this will take up 81 gigs of memory and I simply don't have the capacity to run that. The cloud version of Qwen 3.5 is actually 397 billion parameters, so it really shows how these cloud models are extra large and that's why the results are often a lot better than when running locally.
Step 4: New OpenClaw Installation with Ollama
Part two, we're going to connect up OpenClaw to Ollama. There's two ways to do this depending on whether this is the very first time you're installing OpenClaw or if you already have it installed and want to switch to a local model. Let's start with the simplest way, which is a brand new installation.
Starting on the official OpenClaw website, scroll down to the installation step here, which you can do through the terminal. Copy this line of code and then in the quick launch options, search up terminal.app. This will open up this terminal application here, which you can paste that line of code into. When you hit enter, it'll start the installation for OpenClaw. This will run in the background for a couple of minutes and once it's ready to finalize, it'll ask you to start the setup process.
Confirm the first question here, I'm just going to prompt in yes, I'm using this on a personal basis. Then after that, use the quick start method. We're going to now select our local model. You can see here lots of models from things like OpenAI and Anthropic, but we're going to scroll up here to the Ollama option. Next, you'll want to confirm your IP and port for Ollama. This is the default port, so you don't have to change that. Finally, you'll select the Ollama mode. Since I only want to run local models, I'm going to select 'local' for this option.
Finally, you'll need to select the model we'll be using for OpenClaw. As you can see, I've downloaded a few models here for Ollama, but if you're doing this, you should just see Qwen 3.5 latest. So, let's utilize that.
The next few steps are just to set up OpenClaw as per usual. This includes setting up a messaging channel as well as setting up maybe a search provider and some skills. And if you want to learn a little bit more about these, then I've got an entire different video specifically for it. But for this example here, I'm just going to skip past most of it just to get OpenClaw up and running so that you can see it in action.
For our final step, we'll select whether we want to use it through the terminal or through the web UI. I personally like the web UI, so I'm going to select that and there we go. We now have OpenClaw up and running. Let's test out how this works with our local open source model. There we go. Looks like it's responding.
From this point forward, you can interact with OpenClaw as per usual asking it to set up routines as well as skills and start doing cron jobs so that it's doing different activities every minute, hour, or day. I have a pretty cool video about the top 10 ways to utilize OpenClaw as well, which you can check out here. But before we get further,
Step 5: Switching an Existing OpenClaw Setup to Ollama
let's actually have a look at the second way. In case you have OpenClaw already installed and you now want to swap it from a cloud large language model to a local model running on Ollama, the steps will be more or less the same. I'll need you to install Ollama, but the next part will be to change its configuration in order to connect up to the local model. Let's do that now.
On the Ollama website, head over to the model section and select the model that you want to connect to. This should be a model that you've already installed. In this example, I'll use Qwen 3.5. Down here, you have some application settings. We've got one here for OpenClaw. I'm going to copy out this command here that we can now paste into our terminal. Let me paste this in.
With that done, that's pretty much it. This is the setup that is already pre-configured inside of Ollama to connect up to OpenClaw. Let's now test out OpenClaw to make sure that it's all working. And it looks like that's working both in the terminal and the web UI. If you ever want to check which model you're currently using, you can head to the very top here in OpenClaw to have a look at those.
Step 6: Integrating Apps with Zapier MCP
The final thing I want to show you is how to connect up your apps and services directly to OpenClaw, but through a middleman. A lot of people are connecting them through APIs, but one of the best ways to do so is through MCP. And for this, I'll be using a Zapier. Here's how to connect up OpenClaw to Zapier. Simply ask it to. Just ask how to configure Zapier MCP. It's going to give you some instructions and let's do those now.
We'll want to head to this link, which is mcp.zapier.com. Then you want to click to create a new MCV server. Scroll down and select the option 'other'. Right now, I just want to do some testing, so I'll only connect it up to Gmail to find emails and maybe also to create drafts. I don't want it to start deleting or sending emails on my behalf. This is the benefit of being able to have a little bit more granular control over the types of apps you give OpenClaw access to.
Now to establish the connection, I'm going to generate a token and rather than me manually configuring this like a programmer, I'm just going to copy paste all of this content straight into OpenClaw and hope that it sets it up all for me. And after a few seconds, here's the response. It's all configured and it's also got the available tools I've given it access to such as getting emails and creating drafts. Let me test that this is working by getting the latest five emails from my inbox and here they're presented in a nice chat layout.
Conclusion and Next Steps
I hope you guys enjoyed this video. I definitely recommend checking out my use cases for OpenClaw as well as my crash course if you want to learn more about it. Additionally, if you want me to make more OpenClaw videos, let me know what topics to cover. Finally, a big thanks to Zapier for sponsoring today's video. If you want to learn more about them, definitely check them out in the link in the description below.
