Transcription
A while back, the team behind Trey AI released a powerful open-source multimodal AI agent stack called TARS, developed by Bite Dance. Agent TARS is a core multimodal AI agent stack which brings the power of GUI agents plus vision into your terminal, computer, browser, and products. It ships with both a CLI and a web UI, which is going to let you work with multimodal large language models in a way that feels closer to humanlike task completion. It also integrates seamlessly with various MCPs. It also works with operating across real-world workflows, which is pretty cool.
For example, TAR was asked to book the earliest flight from San Jose to New York on September 1st and the last return flight on September 6th using Price Line. The agent handled the entire process autonomously from searching flights to selecting the correct options, showcasing its ability to function like a digital assistant that truly gets work done across apps as well as browsers.
But then you also have UI Tar's desktop. This is another toolset, which is a native desktop app that provides a full GUI agent experience. Think of this as an AIOS that's capable of operating both local and remote computers, as well as browsers. It ships with specialized browser operators and system operators to carry out tasks end-to-end. Together, these two tools form a powerful multimodal workflow system which enables humanlike AI task execution through cutting-edge models and seamless tool integrations.
To reiterate this, in essence, agent TARS is the general multimodal AI agent stack that works within the CLI, as well as having a web UI for usage. And then UI TARS, on the other hand, is a desktop app built on top of the TARS stack.
We can first get started with agent TARS. And I'll leave this repo in the description below, but essentially, we're going to be installing this within our CLI. And it's super easy to get started, especially if you're having this integrated within your terminal. The only prerequisite is having Node installed. Make sure you have the latest version, as they have stated over here. And then you can run it with your preferred model provider. I have already set my key for OpenAI, but you can follow this command to integrate it with Bulk Engine, which is something that they actually recommend. All you got to do is launch it with the npx command. It is super simple. Just make sure you have the npx command installed. And once you have opened up your command prompt, you can simply go ahead and paste in this command, which will launch agents, and it will install it onto your terminal. And you can see that agents is now available within our CLI. You can obviously use it through the CLI, or you can access it through the web UI, which you can access through your localhost. So I can go over to this localhost, and this is where I can actually start accessing agents with this web UI.
And now, within our localhost, you can see that it is using the browser agent to search different sorts of things. I simply just sent it a prompt to tell me the top five most popular projects on Product Hunt. And you can see right away, it provides me the top five popular projects, and it's able to do it pretty quickly. So this is essentially a great multimodal agent that is going to be used for various sorts of things like tool usage, having it execute browser-based tasks. In this case, I simply just provided my prompt to go over to my channel and provide a summary of it, then write a report. And you can see that it is actually doing this by talking about what my channel is about, AI tools for no-code development, open-source AI agents, etc., etc. And now it's working on the report. In essence, you would use agent TARS as a multimodal agent that would be used to automate real-world tasks like browsing the web, system commands, app control, or even just handling different sorts of data by combining the GUI agent as well as the CLI tool.
Now, with this web UI, you can also execute terminal-based commands. So in this case, I just simply asked it to install cloud code for me, and then it was able to install the packages directly within my CLI, as TARS also allows you to use MCP. So this way, you can actually have it so that you can integrate something like a server chart MCP, and this is where you can actually get visualizations using the MCP to display your data. Now, in other cases, you can use other sorts of tools like a mapping system, and you can essentially have it draw charts, have it basically represent your data through different sorts of visualization methods. And there's many ways you can do this. Just head over to the config file, and then you can input your MCP server there.
Something I forgot to mention is that these two toolsets are under the Apache 2.0 license, which is good to see. But the next toolset is UI TARS, and this is a native GUI agent that runs on your computer. It's powered by the UI TARS framework, as well as the Seed 1.5 and the 1.6 multimodal models, which is going to enable humanlike control of different local apps that you might have, remote systems, as well as even your browser. This is something that I prefer, and in a way, I believe it to be an AI operating system.
Before we get started, I just want to mention that you should definitely go ahead and subscribe to the World of AI newsletter. I'm constantly posting different newsletters on a weekly basis. So, this is where you can easily get up-to-date knowledge about what is happening in it. Go ahead and subscribe. It opens the autosave feature up within it. This is where you can change and configure settings of your local apps, which is pretty insane. And it's going to be able to essentially automate different sorts of tasks for you locally.
To install this, you can easily head over to the quick start MD file, which I'll leave a link in the description below, and it showcases how you can install it with the model that they recommend with macOS, Linux, as well as Windows. This will require you to actually use their base model with UI TARS, and it works best on macOS at this current moment, but you can easily install and select the UI TARS 1.57B model directly from the app, and then you can start using the computer use OS.
Here is agent UI taking in this task to search and collect the Nvidia stock price, and essentially, it was requested to create a Python script for processing the data, and then it was requested to create a research report on the stock and its analysis it has found using the Python script, and you can see it does a really quick job in doing this, and it executes everything through your computer, as well as while searching the web.
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.
But that's basically it, guys, for today's video on TARS. These are two impressive new tools that have been released by the Bite Dance company. And essentially, it is really dynamic, 'cause you get a lot of different ways to access it. Whether that is through an application, through the terminal, or through the web UI. There's a lot to this, and they are constantly updating it. So, it's a pretty cool tool that you can get started with right away. And it's super easy to install.
But with that thought, guys, I hope you enjoyed today's video and you got some sort of value out of it. I'll leave all these links in the description below so that you can easily get started. Make sure you go ahead and take a look at our second channel. Make sure you join the newsletter, join our private Discord, follow me on Twitter, and lastly, make sure you guys subscribe, turn on the notification bell, like this video, and please take a look at our previous videos because there's a lot of videos that you can definitely benefit from. But with that thought, guys, have an amazing day, spread positivity, and I'll see you guys really shortly. He suffers.