How to connect Ai Agents to ANY software | No code

In this video, we're going to be talking about some ways that you can integrate pretty much any software you like with your AI agents, because depending on the AI agent build, you might need to integrate with lots of different softwares. This could be things like CRMs, email accounts, other automation platforms, or other LLMs. There's going to be lots of cases where when you're building these AI agents, you're going to need to integrate with different softwares. And in this video, I wanted to show you really, really simple ways that you can integrate with other softwares. If you don't know who I am, my name is Rory Ridgers and I build AI agent systems for businesses. So that's something you might be interested in, then I've left a link to my website in the description and you can book in a call with me, or you can just fill out the contact form and we can talk about some ways to integrate AI systems into your business. In this video, I've created an AI agent with 5 different tools, and each of the tools uses a different way to connect a software to the agent. And I feel like I've pretty much covered all bases, and after watching this video, you should be able to integrate any software into your AI agents. So in this video, we're going to be using Relevance AI to make the agent, and then we're also going to be using Apify and Make.com. And if you haven't heard of those, Don't worry, I'll explain how they all work later. So this is the agent on Relevance AI. If we hit edit agent, you can see the different tools that we have for this agent. We have the email generator tool, the Perplexity tool, search Airtable data, Instagram scraper, and the Google Sheets tool. And now these tools might seem quite random and they actually don't really work together. This isn't an agent that you'd really use in your business, but I've just made it to showcase how you would connect different softwares to one agent. So you wouldn't actually use this agent for yourself or in your business. But with these 5 tools, you can pretty much work out how to connect any software to your agents. So we're just going to kick it off straight away. We're going to look at the email generator tool, which is probably the most simple tool out of all of them. If we go to build, we can see how this tool works. So we have the user inputs, we've got user email request and the email recipient. And basically all this tool is doing is just sending an email for you. So you just say to it, I want to send an email to da da da saying 'Hey, you still up for that meeting?' And then it will just send an email saying that to that person. So really, really simple tool. As I said, the user inputs are going to be the user email request, and this is going to be what the user wants to say in their email. And the email recipient is going to be who they want to send this email to. So for the first two steps of this, this is optional, you don't have to do this. I wanted to do it so people could just say to the agent, 'Send an email to Rory,' rather than actually writing out the email address. So what I did for that was made it so that the user input is just the name of the person And then we have a knowledge base set up here. I can show you it. So all this basically is, is just a PDF with lots of names here and then fake email addresses. All of these are just fake apart from my one at the top, is the only valid email address. So we're just going to be using mine for the example. So that's all that is. The first step of that, we'll just return that entire PDF with all the names and all the email addresses. And then we simply use an LLM step to go through that data and then look at the email recipient that the users put here. And then basically just find the email that matches up with the name. So if the input here was Rory, all of the data from that PDF would return. The LLM would look for it, find the name that says Rory, and then find the email that matches that, and then just output the email. So we get the email address for the recipient. And then we have another LLM step here. Again, this is kind of optional, but it basically would just expand on the email request that's put in this input. So if you said something like, are you still up for that meeting, this LLM would just look at that and kind of just expand on it a little bit. It'll say something like, hey 'Hey Rory, just wondering if you're still up for that meeting. Thanks.' You know, it won't go crazy, and obviously you can customize this, but it just expands on it a little bit so you don't have to type out the actual full email that you want to send to the agent. You can just sort of say the general gist of the email. So that's what that step does. The LLM step here will literally just look at the answer from this and then generate a subject line. Really, really simple. That may be something like 'Meeting update' or something similar to that. And then here comes the actual integration. On Relevance AI, it's really, really easy to integrate things like email accounts onto your tools and in your agents. So all you do is you click the plus sign here, go to Gmail, and then in this case it's going to be Send Gmail Email. You click on that and then you literally just sign in with your Google account and then boom, you're connected up. It's really, really easy. The to email is going to be the answer from this first LLM step that was finding the email address, and then the email subject is going to be the output from this LLM, and the email body is going to be the output from this LLM. So really, really simple. I'm going to run this through one time just to show how it works. So I've put, are you still up meeting tomorrow, and the email recipient is going to be Rory. We'll hit run tool. It should do this easily. Okay, so you can see it's returned all the data from the PDF here, and then the LLM has gone through that and outputted my email. This LLM step here is just expanded on the email body that we put a little bit and said, hey, still up for that meeting tomorrow? Just wants to confirm if we're still on. Looking forward to our discussion. Best, Rory Rogers. And then this LLM step here has just looked at that and generated a really simple subject line. Confirmation for tomorrow's meeting. And then as you can see, this step looks like it's worked and it should have sent us off an email. Then yeah, if we check my emails, you can see confirmation for tomorrow's meeting. Hey, still up for the meeting tomorrow? Just want to confirm if we're still on. So perfect. That's a really, really simple tool. And Relevance AI make it really, really easy to connect with their native integrations on Relevance AI, things like Gmail. So that's really simple. The next one I'm going to show you is also quite a simple one. This is Perplexity Perplexity tool, and the only inputs here are going to be the user query. So this is going to be just what the user wants to say to Perplexity. And then here we have an LLM step, and all we do for the prompt is just put in the user query from here. And the reason why I put this into this tutorial is I've done it as Perplexity, but you can do it as any LLM you want. So it doesn't have to be Perplexity, this is more just how to integrate with any LLM. So if you look, you can see we've got all the Claude models here, obviously Perplexity We have things like Gemini, Groq. You can even look up Deepseek if you want to connect to Deepseek as well. So pretty much any LLM you want, you use this really, really simple tool to be able to do that. So again, we can just test it out just so you know that it actually works. Obviously Perplexity is kind of like a Google AI type thing. So you pretty much ask it questions about anything and it can take real-time data from the internet and answer your questions. So we're going to do that. So we said, what's the weather like in London today? This will send this off to Perplexity and it should give us an answer back. Okay, yeah, really, really simple. Current conditions, it feels like temperature, the humidity and wind, precipitation, everything like that. Obviously this is really, really simple. And as I said, this is for any LLM, so you can connect any LLM to your AI agents, which is really, really cool. Now we're going to look at searching an Airtable database. So if we have a look at this tool, for this one you can't natively do this on Relevance AI, so we're going to have to connect to a Make.com scenario. But you'll find if you're using Relevance AI, you have to do that quite a lot. So it's a really, really important skill to learn when you're connecting to a lot of external software. So user query, this is just gonna be what the user wants to get from the Airtable database. I'm gonna show you that Airtable database so you know kind of what I'm talking about. So this is the Airtable database we have. It's kind of just demo data on clients that a business might have. So you can see we've got the names here, the business name of this client, some notes, a status, and then a website URL for the client. So really, really simple. Obviously a lot of you guys might use a CRM for this kind of thing, but I just wanted to show you how to connect to Airtable because I feel like Airtable is a little bit more diverse. You might have some other stuff on here, so I feel like it's a bit more valuable to know how to connect to Airtable rather than just a regular CRM. So if you look at the tool again, we've got the user query, so what the user wants to find. This could be something like, what is the website URL for Michael Carter's company, or something like that. So that's what we're going here. And then we send off an API request. And this is where we're going to be connecting with our Make.com scenario to be able to actually talk to Airtable. So I'm going to show you that scenario now. So this is the Make.com scenario. It's really, really simple. And if you've never used Make.com before and you don't know what you're looking at, basically what we have here is we start off with a webhook, and this is how we're going to be able to send data from Relevance AI to Make.com. So if we look at Relevance AI, we have the API request here. The method is going to be POST, and then the URL Really, really simple. It's literally just the URL that we get here from this webhook. So when you put a webhook down, you get to add a name, things like that, and then you'll get this URL here. You're just going to copy that URL, put it into here, and then the first time you use it, you'll have to hit redetermine data structure and then run this API request, and then they'll be accepted. So they'll be linked, and then you can pass data to and from the Relevance AI tool and the Make.com scenario via this webhook. So Really, really simple and very, very useful. This Make.com scenario gets triggered by this webhook, and then we have a simple Airtable search record step. What we do is we hook up our Airtable database. So for this one, it's just called Business Data and then Table 1. And then you can see the output fields here, which is going to output all of the data from this table. So we want the name, the business name, the notes, the status, and the website URL. And then what we do, so we're using an array aggregator. And we select the variables that we want. So again, the name, the business name, the note status, and the website URL. And it's just going to bundle all those data points into an array so it's much easier to read later on for our LLM steps. And then we use a webhook response, and all this does is it sends data back to the webhook that triggered this scenario. And what we put in here is just the array from this array aggregator. So that array is basically just going to be all this data here. We're going to send that through the webhook response back to our tool in Relevance AI. And then we have an LLM step here that's going to go through that data and basically look at what the user's query was and then try and output a response to it to help them out. So, so you can see I've given an example here. If the user's query was, what is the update on the Stellar Studios job? So if we look at the business data table here, Stellar Studios is here. Then we want the LLM to output the status of the data that is associated with Stellar Studios and also maybe the the notes if they are relevant. And then we've just given it another example here, but we're gonna test this out to show you how this works. We're gonna pick one of these clients. We're gonna say this one, Urban Adventures. We're gonna say to the tool, what is the update on the Urban Adventures job? What is the update on the Urban Adventures job? Let's make sure to run the scenario manually on Make.com. Obviously this will be set to automatic, but just to show you guys, We're going to run it manually, and then we're going to run the tool on Relevance AI. You see this API gets sent off really, really quickly. It returns it well quick. So if we just have a look at the Make.com scenario here, so it got triggered by the webhook here, and the Airtable node got triggered as well, and that outputted all of the data from that table. And then the array aggregator bundled that into nice neat bundles so that the LLM can read it. And then the webhook response just sent that entire bundle back to the tool on Relevance AI. And then if you look at Relevance AI, you can see we've got the table here. So pretty much just an exact copy of this table, which is really nice. We've got that table, and then the LLM step went through that data. This is what it output to us. It said, here's the information regarding the Urban Ventures job. So status is prospect, the notes are interested in integrating booking functionality into their current site, and then this is their website URL. I just realized I said Urban Adventures before, it's Urban Ventures. Sorry about that. But yeah, again, really, really simple. And these Make.com scenarios are going to be golden when you're using Relevance AI to make your AI agents, because on Relevance AI there aren't a lot of native integrations. You can't natively integrate with Airtable, which is kind of annoying. So these Make.com scenarios come in really, really handy, and they're really, really simple to set up. Now we're going to have a look at the Google Sheets tool, which is fairly similar to the Airtable tool in that we're going to be using a Make.com scenario. But the reason I want to show you a Google Sheets tool is because On Relevance AI, you can't integrate with a lot of the Google Workspace, which a lot of you guys probably will want to do. I want to all the time, so it's quite annoying. You can integrate with things like Gmail and Google Calendar on here, but that's really about it. So I want to show you how you can use anything from the Google Workspace in your AI agents. So again, the input is going to be user query, and this time we're almost doing the exact same thing as the Airtable database, but we have a Google Sheet set up. I'll show you that, it's kind of different. So this is a Google Sheet, and this time we just have bunch of different products. So we've got the product IDs, the names, the categories, the price, the size, and the color. And this is obviously just example data that an e-commerce brand might have. So you can see we've got like 50 different products in here, and we're going to pretend that we might want to retrieve some data from those Google Sheets. So similar to what we did with the Airtable, we might say something like, how many products do we have in a size extra large at the moment? And then we want the AI agent to return all of the products that we have in extra large. So the way we do that, again, we have the user query, and this would be something like, what products do we have in extra large? And then we send off an API request to another Make.com scenario, which I'll show you now. So this might look familiar because it's literally almost exactly the same as the Airtable one, but we have a webhook here and that connects to this API step. And then we have a Google Sheets search rows, and again, we've just hooked up our sheet and we're just going to return all the data from that sheet bundle it all together with an array aggregator, and then use a webhook response to send that back to Relevance AI. And then same thing again, which can again, LLM step to go through all of that data from the Google Sheets and then look at the user's query and try and answer it. By the way, I didn't mention this, to connect up your credentials on make.com, things like the Google Sheets, it's really, really simple. All you do is you go to connection, click add, and then if you have an @gmail account or something like that, you might need to go into the Google Cloud, set up a new project and get an API key. But there's docs on make.com that make it really, really simple on how to do that. It's quite a simple process. But if you don't, if you just have like me, an autonomously.ai, so a domain that you've bought, then you can literally just one-click integrate. It's really, really simple. But if you don't, which a lot of you don't, if you just have like @gmail or @outlook or something like that, you will have to set up a project on the Google Cloud But like I said, Make.com make it really, really simple. They've got loads of docs and they walk you through basically exactly how to do it. So we're just going to test this out. We're going to look at the data here and we're going to say something like, what products do we have for under $50? So I said, what products do we have for under $50? So make sure to run this tool again. Obviously this will be set up so that it's automatically running, but we're just going to run it manually so you can see how it all works. Run the tool on Relevance AI. If we look on make.com scenario, this gets triggered, gets triggered by the webhook here that's coming through from Relevance AI. The Google Sheets search row will just return all of the data from that Google Sheet. The array aggregator will bundle it all together into— should be, yeah, 50 different bundles. And then we send that back over to the Relevance AI tool. Perfect. So then the LLM step went through this data that we got here, and then this is what outputted to us. Here are the products available for under $50, and then you can see these are all the products obviously under $50. So Really, really simple. I just want to show you guys how you can integrate with the Google Workspace because it's really important and it's really handy. You're probably gonna have to do it all the time. Things like Google Sheets, Google Calendar, Google Docs, Google Meet, everything like that, you can integrate with all of them through Make.com. And now for the Instagram scraper tool, we're going to be using Apify, which I'll show you in one second. With a lot of these social medias, depending on what you want to do with them, you probably can't do it on Relevance AI. You might be able to do it on Make.com, and you also might be able to do on Apify. If you want to scrape social media, so you want to get all the Instagram profiles that have under a certain amount of followers and are in your niche, for example, then you'd have to use Apify, which is best for social media scrapers and kind of any scrapers in general. But if you wanted to do something like post content to your Instagram account, you do that through Make.com. But as I've already showed you Make.com, I'm going to show you how you can scrape social media using Apify. So for this tool, the input is going to be company name, and this is assuming that you want to get some data on a particular company's Instagram profile. This works best with companies because I'll show you why now. The way we get the Instagram profile name— we need to have the Instagram profile name— is we just do a Google search for this company's Instagram. So this could be something like Gymshark. We do a Google search for Gymshark's Instagram, and then we have an LLM step here that's going to go through those search results, look at all the URLs, find the URL that's most likely the actual link to that company's Instagram. And it's going to extract the username from that URL. So as you can see, if the URL looks like this, then the name would be Gymshark here. So it extracts this part of the URL to get the exact profile name because we need the exact profile name, send that off to Apify. Now that we have the profile name, we use an API step, and this is where we're going to send data over to Apify. So I'll show you that now. So this is Apify, and if you've never heard of it before, it's basically just this API store for scrapers. It's really, really useful. As you can see, we've got the Google Maps scrapers, Facebook scrapers, Amazon scrapers, TikTok scrapers, everything like that. In this video, we're going to be using the Instagram profile scraper. So if we click on it, this is the only bit we really need to worry about for now, which is the input. You can see it's got manual inputs, but obviously we don't want to do that because a manual input would be if you're using it on Apify, but we want to use it in our AR agent. So what we do is we go to JSON And this is just the JSON object that needs to get sent through to this actor. They call them actors on Appify. Needs to get sent to the actor in order for it to run properly. So it's really, really simple. It's literally just the usernames and then the actual username. So this is what we sent over, the usernames and then the LLM answer variable, which is going to be this answer, which should be the name of the Instagram profile. So, so the method is POST. The headers are obviously going to be Content-Type application/json because we're sending over a JSON body. Body is going to be JSON object, and then as I showed you, this is what we're actually sending over. And then the URL, the way we get that is we go to our actors on Apify, go to API, API endpoints, and then we scroll down to run actor and get dataset items, because this is going to run the actor and then wait for it to finish and actually get the data that it retrieves. If you do some of these other ones, if you just do like run actor, then the API set will just run the actor and it won't actually wait to get the data that it retrieves. But in this case, obviously, we want the data. So we use this one, which is going to run the actor and then actually wait for the data to get back to us. So once we have the data on this Instagram profile, we're going to use an LLM step to basically just look at that data and format it nicely for us so it's a little bit easier to read. So we're going to try this out quickly. We're going to use the example that I used before, Gymshark, just because it's an easy one. We can hit run tool and then we can see this in action. We can actually see the actor working on Apify. Which is quite cool. So these two steps have been triggered. The Google search, we got this, and then the LLM step obviously went through that and outputted Gymshark, which is the name of Gymshark's Instagram profile. The actor, we can go to runs and then see, it's this one. It's obviously worked. It did it quite quickly, which is nice. This is what it gave us. So if we look back on our Relevance AI tool, this is what it outputted to us. I'm not going to click view all because for some reason the actual data is right at the bottom of this long table and it's really long. I don't know why it does it like that, but but it is there, I promise. You can see the different headers here, the URL, the ID, the username. And then we use this LLM step here to basically just output this back to us. And this is in markdown format, so it might look a little weird now while we're in the tool, but when we're talking to the agent, the agent will just convert it into plain text for us so it actually looks normal. So, but if we look at the actual content here, you can see we've got the Instagram profile, Gymshark username, the full name, the bio, their URL, their follower count, following, post count, Verified is yes, business account yes, and then their highlights, they've got 12, and some of their related profiles. So really, really cool. There are a lot more powerful actors on Apify. I just want to use a really, really simple one to show you guys how it works. Can't get more information from these profiles, but the actors take a little while to run and they're a little bit more complex to set up. I just want to show you a simple one because I feel like if you can connect an agent to a software on Relevance AI, if that's what you're using to make your AI agents Make.com, and then also Apify. You can pretty much connect it to any software and get to do pretty much anything. So that's why I want to show you just a simple way of using Apify. But now we obviously have to put all of these into an agent. So that's what I showed you at the start of the video, the multi-use agent. If we hit edit, we've got all the tools here that we've just been through. The core instructions— this is really quite simple. We have the role: you're a helpful assistant that's able to use a variety of different tools to be able to fulfill a user's request. The task is you have 5 different tools that don't really work in conjunction with each other. However, they each have their own use and help a user, because as I said at the start of the video, these tools don't really work together. It's more just an example for you guys to show you how you can connect with different softwares. So that's why I said that to the agent, just to make it clear to it that these tools don't really work together. It's more of just an example. Your job is to identify what a user is trying to do based off their input to you and then use a correct tool. I'll give you a breakdown of each of the tools now. And then for the tools, as in every video, I pretty much follow the exact same format on how I prompt the agent to use these tools. I give it the name of the tool, and if you don't know how to do that on Relevance AI, you can just hit a forward slash, go to tool, and then directly reference the tool like that. Just makes it a little bit clearer for the agent. We say to it what this tool does. So this tool allows users to search a Google Sheet containing product info, and then we give it the user inputs. The only user input is query. And then we tell it what to put in these inputs. So put the entire user's query into this input. And for these tools, I'll either tell it when to use these tools or, or I'll give an example of what a user might say in order to trigger this tool. In this case, I've given an example. So I've said, I want you to use this tool when a user requests info on a product. This could look like they use the same, what products do we have in a size small? Or how many products do we have left in blue? And then I've pretty much followed the exact same format for all of these tools here. As you can see, you can pause the video, read through it. But it's basically just the exact same format. So what we're going to do is we're going to hit Save Changes and then test this agent out and see if all these tools work while they're in the agent view, because we know that they work while they're in the tool view, but while they're inside an agent, will they work the same? So we're just going to work our way through these tools. We're going to start with the Google Sheets one. So I've just said to Matt, how many black products do we have? So obviously we want it to look through that Google Sheet on all that fake product information and can give us back all the products that are black. So you can see Use the Google Sheets tool. These are 6 Black products available. Here are their names. And it's given us the names of them. So absolutely perfect. We're just going to ask it a question that it might have to look up using Perplexity, something like, what is an LLM? So kind of something that the AI agent might not know off the top of its head. We'll have to use Perplexity to look that up. Let's have a look. So yeah, it's using the Perplexity tool, which is perfect. Okay, so this is what it's given us after it's used the Perplexity tool. Master the laws, large language model. It's given us a training architecture, word embeddings, capabilities, application. So yeah, pretty pretty cool. I'm not going to read all through that, but that's obviously what an LLM is. So perfect. Now we're going to try and use the email generator tool. Let's say, send an email to Rory saying I'm going to be late for the— so send an email to Rory saying I'm going to be late for the meeting. Should use the email generator tool. Let's see. Yeah, one step performed in the background using the email generator tool. Perfect. Okay, it's used the email generator tool and it said, I sent the email to Rory informing him that you'll be late to the meeting. Now if we look at my inbox, hey, just want to let you know I'm going to be late for the meeting. I apologize for any inconvenience this may cause and I appreciate your understanding. See you soon. Best, Rory Ridger. So absolutely perfect. It's really, really easy, that one. Now we're going to try and use the Instagram scraper tool. Let's say I want some more info on— so I've said I want some more information on OpenAI's Instagram. Should use the Instagram scraper tool. Let's have a look. Yep, using the Instagram scraper tool. Perfect. We can also look at the inputs that the tool uses for this. We look at the input for company name, it's OpenAI, which is obviously perfect. Okay, and this is what it's given us. Here's some information about OpenAI's Instagram profile: their username, their full name, their bio, their website link here, their followers, following, post count, highlights, business category, verified, and some of the related profiles. So absolutely perfect. Obviously use the actor on Spotify. No problem at all, that's perfect. And now we're going to try and get to use the last tool, which is the Airtable tool. We said what clients are looking for website builds because that demo data is from sort of a digital marketing agency. So I said what clients are looking for website builds. Let's use the search Airtable data tool really, really quickly, which is really cool. And then let's see what outputs to us. Okay, so here we go. Here are the clients who are looking for website builds along with their relevant details. So John Doe is interested in website redesign. Tom Harris wants website. Michael Carter signed a contract for a portfolio website. David Wright, Emily Green. And then the total clients looking for website builds is 5. So perfect, they all work absolutely perfectly. So we managed to connect 5 different softwares to our AI agent using different ways to connect all of them, to show you guys how you can connect pretty much any software you like to your AI agents. So if you have any softwares in mind that I didn't connect to in this video, your best bet is probably trying to connect to them through make.com or maybe Apify, or just looking on Relevance AI to see if they have any native integrations for that software. As I said at the start of the video, if you were interested in working with me so we can integrate some AI systems into your business, then I've left a link to my website in the description. You can go ahead and book in a call with me, or you can just fill out the contact form and I'll get back to you. Hope you guys enjoyed this video, and if you have any video suggestions that you'd like to see, please do just drop them in the comments and I'll try my best to get around to them. Make sure you subscribe and stick around because I've got some great videos coming out in 2025.