Unlock the Power of External MCP Servers in Your UiPath Agent Today

In this video, we're going to use an external MCP server from within a UiPath agent.

[Music] So, in my last video, what we did was we created a UiPath MCP server and consumed it from within Postman. Today, we're

kind of doing the other way around.

We're going to connect to an external MCP server inside of UiPath and then in an agent project in Studio Web, we're going to use the tools that that MCP server exposes. So the first thing we're

server exposes. So the first thing we're going to do is we're going to go to Smithery AI. Smithery AAI is a website

Smithery AI. Smithery AAI is a website where you can find skills that you can use in your AI projects. So I just happen to know that if we search here

for meow human we get a list of of different uh things that this user has published. If we go to weather here this is an MCP server that exposes two tools. Get alerts which

is basically a tool that lets you get weather alerts for a certain um US state. You need to provide the twolet

state. You need to provide the twolet state code and then get forecast which will get you a weather forecast. But you

need to provide it with a latitude and a longitude. So a set of coordinates and

longitude. So a set of coordinates and if you do that it'll give you a weather forecast. Now let's add this um MCP

forecast. Now let's add this um MCP server to our agent project. So inside

UiPath Studio I have just started a new agent project. In my environment I have

agent project. In my environment I have access to some preview features and one of them is conversational agents. This

is a pretty cool thing and it's coming and I want to show it briefly to you in this video. So I'll just select that,

this video. So I'll just select that, click start fresh and uh what we can see is basically we have a completely empty um agent project here. What we want to

do is we want to add an MCP server to this project. Now we don't have any MCP

this project. Now we don't have any MCP servers in our UiPath setup. So we're

going to add that first. I'll jump into orchestrator, click my demos folder, click MCP servers, and we can see that we don't have any MCP servers here. I'll

click add MCP server and then we can see that there are different uh types of MCP servers we can add. In the first video we worked with a UiPath MCP server. Here

we're going to work with a remote MCP server and we'll just give it a name.

We'll call it weather service like that. We can also give it a

like that. We can also give it a description. Let's

description. Let's us get weather information. Just a description.

weather information. Just a description.

And then we need to provide it with a remote URL. This is the URL for the MCP

remote URL. This is the URL for the MCP server. So we'll go back to Smithery and

server. So we'll go back to Smithery and we'll copy this URL that we have here.

Copy that. Go back into uh orchestrator and add it down here. Now we do need to provide an API key. So I'll add uh an

argument to our URL here by typing in a question mark and then type in API_key equals and then we need to type in the

uh API key. In Smithery AI, you can get an API key, but you do need to sign in first. I've signed in with my Google

first. I've signed in with my Google account, and the way I get the MCP server for this uh the API key for this uh MCP server. The easiest way is just to uh go to the JSON section here,

scroll down, and then you can find your API key down here. So, I'll copy that.

I'll go over to the add MCP server screen again, and then just paste it in.

That's it. And click add. So now we have added the MCP server to our UiPath configuration and we can see here that the tools including their description is

uh is now available inside of UiPath. So

if I go to agent uh builder again inside of studio here. Now if I click add MCP server the weather service MCP server is available. If I click that it's going to

available. If I click that it's going to fetch those descriptions. I'll just

expand this a little bit. It's going to fetch those descriptions and those descriptions are what lets your agent know what the tools can do because it's natural language. So, if you ask for a

natural language. So, if you ask for a weather forecast, the agent is going to look for a service or a tool in its catalog that can give you a weather

forecast. And funny enough, we just

forecast. And funny enough, we just added one of those via the MCP server.

So, I'll click connect to MCP server here.

And now basically what I can do and this is very very simple agent. If I click debug uh the project is going to run and we'll soon see a little chat window

where we can ask our agent questions and hopefully it can answer those questions because it now has access to tools that will help us answer those questions. So

now I can talk with my agent. So, uh

I'll just ask it. Uh what's

the weather like at uh at Pike Place Market? Pike Place

Market is a famous market in Seattle.

I'm not providing it with coordinates.

I'm not providing it with a state code.

I'm just asking for location. So

hopefully the LLM will know that to translate this into a set of coordinates and then check out what tools does it have available. Actually it'll do it the

available. Actually it'll do it the other way around. It'll check the tools.

Do I have any tools that can give us a weather forecast? And it'll find out yes

weather forecast? And it'll find out yes I do have a tool but it needs coordinates. Then it should hopefully

coordinates. Then it should hopefully try to uh translate the location Pike Place Market to a set of coordinates and then call that tool with those coordinates and then give us a weather

forecast. Let's see what happens when I

forecast. Let's see what happens when I click the button here.

So, what it actually found out was that I didn't give it a set of coordinates.

And that's correct. And it needs a set of coordinates. And then it looked it

of coordinates. And then it looked it up. Um, just because it's an LLM, it can

up. Um, just because it's an LLM, it can do all kinds of stuff. And so, it's asking me, are these coordinates acceptable? And I'll just say yes. Go.

acceptable? And I'll just say yes. Go.

and we'll see if it actually does find a weather forecast for for Pike Place Market.

And there we go. It's generating a response for us. So, what we did was we added an MTP server to our UiPath environment that uh sort of downloaded

the description of the tools exposed by that server. We added the server to our

that server. We added the server to our agent project and just by writing out, you know, our goal, we need a weather forecast for this location. It found out that the tool needs a set of

coordinates. The agent could translate

coordinates. The agent could translate our location to a set of coordinates and then call the tool with those coordinates and give us the weather forecast. I think this is pretty cool. I

forecast. I think this is pretty cool. I

think it's pointing in a direction where the way we build automations is going to change not just a little bit, it's going to change dramatically. And the more

tools and the better described the tools are, the more powerful agents we can build with them. This was I know a very very simple demo. I hope you liked it.

If you did, give it a thumbs up the video. And also, if you want to see more

video. And also, if you want to see more videos about agents and stuff and UiPath in general, uh, subscribe to my channel and don't be afraid to share this video with your network. Uh, it would really help me a lot. So, I hope you liked it.

Hope to see you next time. Thank you.

Bye.