The Most Underrated Tool for AI Agents Right Now
Publicado no IA Dots ·
Agentes de IAAI SignalAssistente pessoaldot-ai_agentsdot-artificial_intelligencedot-claudeidioma-enInglêsyt-c2QukPONnfE
Vídeo em Inglês, publicado pelo canal Tech With Tim.
About this content: The summary, key points and topics below are generated automatically from the video transcript. They reflect the content, opinions, claims and predictions presented in the video and do not represent a position, validation or fact-check by IA Dots.
Video summary
The video discusses the limitation of AI agents in reliably retrieving data from the internet due to website blocks and dynamic layouts, and presents Apify as a tool to solve this issue through an MCP server integration with AI tools like Claude, enabling automated data extraction from platforms like Google Maps, Instagram, TikTok, LinkedIn, and YouTube.
Key points
- AI agents struggle to reliably get data from the internet because websites block automated traffic, load content dynamically, and change layouts.
- Apify is a platform with over 80,000 ready-made tools called actors designed to extract data from various websites.
- An Apify MCP server allows AI applications like Claude, ChatGPT, Cursor, and Codex to search, select, and run actors using plain English.
- The video demonstrates using the integration to pull data for Miami dental clinics from Google Maps, Duolingo posts from Instagram, AI tool videos from TikTok, and AI engineer job postings from LinkedIn.
- The integration can also be used to search YouTube videos, analyze content structure, and combine multiple data sources to build custom agent skills for lead generation and outreach.
Topics
AI AgentsWeb ScrapingApify MCP ServerData ExtractionClaude IntegrationLead Generation
Video transcript
Transcript from available captions; it may contain errors.
AI agents have a massive problem and that's that they can't reliably get data from the internet. Now, most websites block them or they prevent them from pulling it. And today, I'm going to fix that with a oneline setup using what I think is one of the most underrated tools for AI agents right now. And by the end of the video, your agent will be able to pull data from places like Google Maps, Instagram, Tik Tok, LinkedIn, and YouTube, and all the other sources that previously were almost impossible. Now, here's the thing that nobody really tells you about AI agents. The models are incredibly smart, but they're almost always stuck inside of a box. Now, watch what happens here if I ask Claude for something simple like the rating and phone number of every dentist in a neighborhood in Miami. Now, in most cases, it either tells me that it can't find that or it searches the web, reads a couple of pages, it gives me a few results with half the details missing. I mean, look right here. You can see we get some of it, but it's really not reliable and we're missing a ton of the info. And that's not because the model here is dumb. It's because the websites that have the data that we want, so places like Google Maps, Instagram, Tik Tok, LinkedIn, whatever, they're not built for AI agents. These type of sites block automated traffic. They load content as you scroll, and they change their layouts all the time. So, normally, if you want your agent to do real research, you'd have to build integrations for every single one of these sites, deal with proxies, and then fix everything anytime one of those sites changes. Now, that's weeks of work before your agent actually does anything useful. So, in this video, I'm going to show you how you can skip all of that with literally a oneline setup that just makes Cloud Code have the ability to do this. Now, the tool that I'm going to use here is called Appify. Now, Appify is a platform for getting data out of the web. And the part that matters the most for us here is the Appify store. So, if I scroll through it, you can see that it has over 80,000 ready-made tools for pulling data from pretty much any website that you can think of. So again, we look here, we have Google Maps, we have Instagram, we have Tik Tok, we have YouTube. All these ones that are notoriously really difficult to get data from. And then if you want to go lead genen, you can pull stuff here. If you want Zillow, you get the idea. Pretty much anything you want is available here. Now, every single one of these tools is called an actor. And an actor is just a small program that runs in the cloud and does one job really well. So for example, this is the Google Maps actor. Now what I can do is press on use actor and then I can just put in a term something like coffee shop and then do the location maybe Austin Texas United States. Go here and press on save and start. Of course I can do multiple runs if I want. Then you're going to see that it starts triggering. Now what will happen here is Appify will scrape Google Maps and give me all the information in a structured format. So you can see I've actually gotten it all back here already. We get all the different results. I can view it directly inside of JSON. And of course, I'm just doing it from the website, but we can actually connect this to our agents, which is where this gets super useful. Now, the idea is really simple. You just give an actor an input, like you saw here. It runs on Apify servers, deals with all of the blocking and the proxies for you, and then hands you back this clean, structured data. Now, you only pay for the results that you actually pull, and there's even a free tier. You can see that it gives you $5 in credits to start, which is actually quite a bit. Then you can upgrade later if you want. And of course, all of the popular scrapers work extremely well. In fact, this one has over 600,000 users. Now, here's why this is so good for AI agents. Well, Ampify has an MCP server. An MCP is the standard way to plug tools into AI apps like Claw, ChatgBT, Cursor, and Codeex. Now, once it's connected, your agent doesn't just get one tool. It can search all 80,000 actors that are available here, read how it works, pick the right one, and then run it. So, you just connect one tool here. You speak with it in plain English, and then you get access to all of the features that you just saw, but from your AI agent. Now, massive shout out to Ampify for sponsoring today's video. I'm going to show you how to connect this and install it inside of Claude and then go through a bunch of demos so you can see exactly how it works. Again, it's free to use. You can make an account from the link in the description. Now, let's dive in. So, once you've created an account with Apify, you can simply go to mcp.appify.com ampify.com and you'll see a list of instructions based on what agent you want to install this inside of. If you want to go with cloud code, make sure you just check add ampify API token. Copy this command in your terminal and run. If you want claude desktop, you can go with these steps. And if you want just claude normally, you can go with these right here. Now, I'll just show you how to add the MCP server directly to Claude because I find that works the best. And this will also work in other tools here. So like Codex, cursor, chatgpt, there's all kinds of connection steps. It shows you exactly how to do it. I'll leave the link in the description. So, for now, I'm just going to copy the URL of this mcp.appify.com. I'm going to go into claude. I'm going to find customize and then I'm going to go to connectors. From here, I'm simply going to go add custom connector and then I'm going to give this the name of ampify and then type in the MCP server URL. From here, I'll press on continue. And it's going to ask me to sign in where I'll need to actually do the authorization and sign in with my appy account. In terms of OOTH client, we'll just leave this as the default option and go ahead and press on add. So from here, it should prompt us to connect to Ampify. So we'll just press on that connect button, which should open up a login screen in our browser where we'll be able to authenticate. So you can see it's brought me to my Appify account. So I'm just going to go and press yes and allow access here. So that Claude will be able to connect. Okay. So let's go back to open claude. And now we should be connected to Appify and it should start loading the tools. And there we go. We see we have interactive tools, readonly tools, write and delete tools. For this video, just to make it easy, I'm just going to go always allow for all of the tools like so. Perfect. So now we've got those ready to go and we can actually start using Ampify and I want to start running some demos. Okay, so for the first demo here, I want to repeat what I attempted in the intro where I tried to get all of the different leads of Miami Dental Clinics using normal claude code or using normal claude. So, what I'll do now is repeat it because we've got Appify installed and we'll see if we get a complete list this time. So, let's paste this in right here. And I'm just going to add a prompt and say, "Can you give me a really nice formatted table at the end?" And just confirm that we have all of the information that we're looking for. So, let's go ahead and press on enter here. And by the way, if you guys are wondering what I'm using for dictation, cuz you always ask, I'm using a tool called Whisper Flow. See, I have nearly 300,000 words. I'll leave a link to it in the description in case you guys want to try it out just because you guys always ask me how I do the voice typing into Claude. Anyways, let's give this a second here to finish and let's see the result that we get. Okay, so it just wrapped up here and this time we've got a full leads.csv file that contains all the information that I asked for. I'm just going to blur the phone number so that we don't have anything being leaked here on screen. Then as you can see here, it says that hey, we have pretty much all the information. The only difference is we're missing a few phone numbers, a few star ratings, and all of this information was just stuff that was missing directly from Google Maps. So, that data didn't exist. It's not like we couldn't find it. It just wasn't there. Now, if I go back to the console here, we can see the full run. We can see it cost us 10 cents in usage. There we have $5 in credit, right? And then I can click into this. I can view all of the results and I can see all the information directly on here. And it's stored in Appify. So, if I ever need it again, I don't need to pay for pay for it. Sorry. I can just pull down the results again. Perfect. So, that's the first demo. And and if we compare it to what we had before, let me actually open that up and we'll see the difference. Okay. So, this is the result that we got before. I actually only had 10 different listings here. I was missing a bunch of other information. I didn't have the ratings and you can see that it wasn't as nearly as complete as what we just got there with the other session. Anyways, let's move to another demo. All right. So, for this next demo, I'm going to paste in another prompt and I'm going to use Appify to pull the 30 most recent posts from Dolingo's Instagram. Now, the reason I'm going to show you this is because typically it's very difficult to pull information from Instagram. That's something that gets blocked a lot and that I've had a lot of difficulty doing using AI agents. So, let's see if that can handle it. Press enter and see the results that we get. All right, so we just got the result back here. You can see it was able to pull all of this information here. Uh, you can see it used ampify to call the actor here for the Instagram and then it pulled Dolingo. We can see we get the reels, carousels, and photos. And then we notice that reals win, but it's uneven, whatever. We get the top five posts and then we could start analyzing this information all from the chat without having to build anything out to go and get that data. Same thing if we go back to the console here, we can go to the runs. We can see that it worked with Instagram. Cost us 5 cents. We got all of the results and we can check it out right from here. Perfect. Now, let's move to the next demo. Okay. So, along the same lines here, I'm going to do something with Tik Tok. So, I just have a prompt. Use Ampify to find the 30 mostviewed Tik Tok videos for the hashtag AI tools. Let's see what we get there in the last 30 days. This is a little bit of a different request where I'm not looking at an account. I'm looking at the hashtags. So, let's see if it can handle that and kind of search more broadly rather than targeted on one profile. All right, so it's just wrapping up now. This one took a little bit longer cuz it was looking at over 300 different videos. You can see that we've got all the different creators. We've got different captions popping up here. And now it's actually looking through or did look through the transcript of those different videos and then dumped it all into this top 30.CSV file. You guys kind of get the idea. And also Claude even went, hey, you know, this is probably related to your videos. Just picked that up from the memory and started giving me some information on what I could do here. Cool. Let's look at another demo. All right, so next let's do something on LinkedIn. Same thing. I've got a prompt here. Use Ampify to get the 50 newest LinkedIn job postings for AI engineer in the United States from the past week. can then tell me which companies are hiring, the salary range, you get the idea. I'm actually curious what this is going to be. Let's run it and see if we can pull the data. Again, LinkedIn, a site that is notorious for being very difficult to pull data from. All right, so just come back with the result right now. We can see that it analyzed 45 unique companies with 50 postings. Uh postings that listed the pay was 78% 150 to 250K makes sense to me. Lowest 99, highest 388 at Tik Tok. Makes sense. Remote roles, hybrid, on-site, work setup not stated. And we get pretty much all of the stats we're looking for. Oh, this one's interesting as well. Top skills. You guys might find this interesting. LMS, Python agents rag PyTorch Evvels NLP, Cloud, TensorFlow. Pretty much what I say in all my videos. Anyways, let's go to another demo. Okay, so we got another one here. This time I'm going to do YouTube just cuz obviously I'm a YouTuber. So I'm going to say, hey, find the 20 most viewed videos on Cloud Code. Hopefully mine's going to pop up there. Publish in the last 30 days. get the title, channel views, channel subscriber count, find the outliers, then tell me how the video is structured, and let's see what we get. Okay, so it just wrapped up here. We can see that we get the table here with the outlier ratio. So, compared to the average views on the channel, how much more did it get? Right? So, like Noah Kim only 5,000 subscribers getting 400,000 views. That's something notable. Some Russian guy here. Okay, I don't really care about that one. Less bidder 22K. Lucas 26K. Money talks with Leon 64K. and then notice all of the views that they're getting. And then it broke it down and explained to me how it was structured based on the transcript of the video, which is pretty cool. Okay, so we've looked at five quick demos. Now, obviously, you understand the capabilities. It works. We can pull the information. At this point, if I go over, we can see we've only spent $2.71, which to be honest is pretty impressive for this amount of data. And now what I want to do is kind of combine all of this together and show you a really practical use case and skill that we could create that would take all of this and give us something meaningful. So let's make a new session and let me show you what I mean. So say you run a social media or AI automation agency and you're looking for clients. Well, the best leads are usually businesses that have great reviews but no social media presence, especially if they're growing businesses. So, I'm going to build a simple command here just by setting up a skill in claude code that can combine all of the different demos that we saw here to research businesses that have strong reviews lacking on social and then write me a pitch that I could send to them so I could start doing some lead generation. So, I've got a pretty long prompt that I just copied in here that essentially tells Claude that I want to create a skill called prospect. Now, the idea is that I want to then give a business type and a city. So, /rospect, you know, Pilates Studio in Miami, whatever. And it should use Ampify MCP and go through the following tools. So, we're going to use all the tools we already looked at. So, we're going to use Google Maps to find the businesses and the reviews. We're going to use Instagram to check how active these are on social media. We're going to use LinkedIn to check who's hiring, see if they have some kind of budget, if they're growing. We're going to use Tik Tok to find what type of content is going viral in that niche, and then YouTube to find what topics in that niche are actually getting views. Then I'm going to have Claude score all of these different leads and write me a personal pitch for the top five that it finds. This is just a simple example. Of course, you can build this out, but let's hit enter here. Let's have it curate this skill and then let's run this skill and see what actionable leads we actually get. All right, so it looks like you finished writing the skill here after looking up the actors we need to use in Amplify. So we've got prospect, we've got the arguments, parse the input, you know, read the rules for the actors. Here's how you do it for Google Maps and you get the idea. Now, because we have this in a skill as well, it means that we're going to save a little bit of time in terms of the tool calls going to Ampify because we already know the structure of the response or the request that we need to send. So, let's just try to run this now. So, we're going to go slash prospect, which is the skill that we just wrote. Okay, so it's popping up there. And then we need a business type in some city. So, for the business type, let's go with, I don't know, maybe barber shops or something in London. And let's see if it's able to look that up for us. and show us what's there. All right, so it looks like this just wrapped up here. It gave us a full report here in this HTML dashboard like I asked for. We can see all the different businesses. We can see what's working in this particular niche. We see all the different content that it pulled. YouTube Barber, most viewed videos from the past month. We have ones with 4 million views. That's kind of crazy. And then we have the pitches prepared. So JNC Barbers, uh Barber Tainer, East London, Culture Cuts, whatever. And it gives us a score on how good it thinks the lead is. And then also the actors that it used, what it pulled, and the amount of money that it spent, and it said approximately 37 cents, which is kind of wild to get all of this information. Of course, I'd have to look through this, fine-tune this skill, make sure that we're improving this over time, but just trying to show you how you could combine this together to do something that's very affordable that could legitimately lead to actually pulling a real client. Anyways guys, that's going to wrap up this video. That's how you use the Appify MCP server. In my opinion, this is one of the most underrated tools out there. It's very cheap and very easy to pull this info into Claude or really any AI tool. Let me know what you're building with it and I'll see you in another video. [music] >> [music]
Assistir ao vídeo original no YouTube
Assuntos: Agentes de IA, Assistente pessoal. Publicado no YouTube em 2026-10-16.
Seleção automática por assunto e idioma. O vídeo pertence ao canal de origem.