Name: Browser-Use
Uploaded: 2025-01-27T11:56:55.304Z
Duration: 2 min 33 s
Description: An open-source Python package enabling AI agents to interact with web browsers for automation tasks.

browser-use

An open-source Python package enabling AI agents to interact with web browsers for automation tasks.

Browser-Use

Looks like these AI browser agents keep popping up left and right. We first saw computer use from Anthropic, which can enable AI agents to take control of your computer and can almost automate anything on it. We had then taken a look at Runner H, which is another version of computer use as it is something that also recorded a higher benchmark score while being faster. 

But today I found something that is open-source and even better than both of those two agents that I've mentioned. It's called Browser Use, which is an AI-powered browser automation framework that is probably the easiest way for you to easily connect your AI agents with the browser. It makes websites accessible for AI agents by providing a powerful yet simple interface for browser automation. 

Browser Use is essentially a new tool for you to connect AI agents to your browser with automated browsing features. You have scraping, interacting with using large language models, and so much more. Just take a look at this demo video where Browser Use is automating job applications. 

In this demo, you can see that the AI agent can, uh, read the CV. It can extract, uh, relevant keywords like Python as well as TensorFlow and use them both to search for machine learning jobs online. In this case, it's also capable of sharing AI Agent Store with colleagues.

Scraping job postings from sites like LinkedIn or even Indeed. And you can basically have it save the results in a structured file. And then have it open the application pages in a new browser tab. And in this case, you can even autofill details for you. 

Now here is another demo video where someone is finding flights on Kayak agent with browser use. In this case, you can see that it's able to input the departure and the destination from Zurich to uh Beijing with this uh designated date. And it's going to be able to then scrape through flight options including prices schedules. It's going to save the data for review or refinement, which you can see on the right-hand side. 

Now this is possibly one of the best functionalities of browser use with this demo of finding a model with a particular license. Now I want you to take a look at this video which will showcase this further in detail. Let us try this next example. Let us go to Hugging Face and find the five most liked models with a specific license. 

So you can see it starts with the Google search cookies pop up. So then damic content and I can just accept them. Now we can see at each step what the model kind of thinks. You see here, okay it sets its next goal going to Hugging Face. You see it already got the sort in.

Correct now it clicked on the right filter. Got the models and now you see it calls this custom function save models. This function is basically just a custom function which we defined here which expects a ptic model with stuff like title URL likes. We could simply extend that here with something else if you want to save more. 

And the agent is now simply capable of calling this function with the specific arguments. Now that was truly just amazing. This is the capability of browser use and from all of those demo videos you are able to see how fast this AI agent is in terms of web based TKS. 

Now if you take a look at this web agency accuracy benchmark test, you can see that the browser use is greatly outperforming web Voyager. You have computer use agent e as well as ronner H. Now this is just insane cuz it's recording an 89 percentage in terms of web agent accuracy which is just near perfect. 

And this is the capability that you get with browser use which is an open-source tool that you can easily install. It's something I'm going to be showcasing as we go further into the video. You have various features like I mentioned at the start. You have Vision Plus HTML extraction, you have multi tab management, element tracking, any large language modal.

Support and so much more. Now if you're looking to install this and get started with browser use, there's a couple of prerequisites that we will need fulfilled beforehand. What you'll need to do is you'll need to install UV to have virtual environment for uh browser use. You need to also make sure that you have Python 3.11 or above installed as well as Playwright. 

Now once you have these three fulfilled, we can then proceed forward with the installation of browser views. So let's get started. I'll leave all the links in the description below so that you can easily get started, but head over to the docs link. Once you are here, go ahead and open up your command prompt. 

What you want to do first is copy this first code to set up your virtual environment, and you can go ahead and paste this in to activate it. Go ahead and paste in the next script which is over here. You can then paste this in, click enter. This will activate the virtual environment. 

You can then install the dependencies within that environment, so go ahead and paste in the PIP install browser use command, and this will start installing browser use within that virtual environment. Once this is complete, you can then install Playwright, and then you can go ahead and create your AI agent, which is super.

Simple to do so. So now I am going to install Playwright. Once that is done, we can then go ahead and start off by creating our agents. 

Now rather than creating your own AI agent, there's actually a couple of examples that you can already use from their template list. These are various templates from uploading files, the Amazon search, as well as various other things. So if you're interested, clone this repository and then you can head over to the example list. You can take a look at most of these examples and you can easily get started with it right away. 

So if you are interested in installing all of these templates, let go ahead and install it by cloning this repository with the git clone command and then pasting in the link. To get the link, go over to this repo, go over to this green button, copy the link to clipboard, and then once you have installed it, you can then go over to the example list and you can start interacting with it. 

Now you can also just simply open up the browser, use the repository, go over to the .env.

Example file go ahead and rename this. Take out the do example and then paste in your API key for your provider. So in this case you can select between these two different providers. Then click save and then you can use any of the templates that we had showcased. 

So I'm testing this out. I have now set my API keys and now I can access the browser use agent. In this case it looks like it's searched up uh Elon Musk and it's a multi uh tab handling task. So in this case it says open three tabs with ilamas Trump and Steve Jobs and then go back to the first and stop. 

So now you can see that it is functioning. It opens up all three. It's going to then go back to Elon Musk it seems and that's how it was capable of executing this task. Now this was something that you can easily get started with and automate anything quite quickly. 

Now let's try something else. We're going to go ahead and have it execute this Amazon search uh python example. We're going to go ahead and have it go to amazon.com search for a laptop sort for the best rating and afterwards it's going to give me a price of the first result. So it's going over to Amazon and it's going to go ahead and execute this command for me. 

Now the cool thing is you can also have multiple agents perform your actions in this.

Case you can see that open two tabs with Wikipedia and it's going to also deploy another agent to analyze the two tabs that were open. This is where you can give a lot of flexibility in terms of having agents do multiple tasks autonomously on your browser which is just insane. 

If you like this video and would love to support the channel you can consider donating to my Channel Through the super thanks option below or you can consider joining our private Discord where you can access multiple subscriptions to different AI tools for free on a monthly basis plus daily AI news and exclusive content plus a lot more. 

Now those were just examples but if you want to create your own agent it's super simple you just need to change your provider whether that's with open AI or anthropic as well as your task. So those two things are essential just need to simply replace the task over here you can do something like searching flights or buying something online or web scraping but you also have the ability to use various other models as well and that's the flexibility that you get with browser use. 

Now this is something I definitely recommend that you try out. Cu you're going to be able to get a lot of flexibility in terms of automating anything on the web and this is.

Essentially the most fastest as well as the most accurate AI agent that is fully open source for you to access right now. So definitely recommend that you try this out and that's basically it for today's video guys. I hope you enjoyed it and you got some sort of value out of it. 

I'll leave all the links that I use in today's video in the description below. Make sure you follow me on the Patreon so that you can stay up to date with whatever is happening in the world of AI. Follow me on Twitter so that you can stay up to date with the latest AI news. 

Note, that there may be slight differences in text and video. Watch video for the original content.

Browser-Use Video