Simple AI Agent Automation with Claude Code and SER API

Okay, so today I wanted to share my super simple framework of how you can kind of use yeah you can use cloud code, you can use codec to kind of automate in a simple way whatever you want. And for this I just have kind of three things I want to use here. I want to use a chron job for timing. This is of course important if you want to have this triggered on a schedule. uh like

yesterday's video, I want to use the claw--p uh for headless and sometimes we really want to use uh the browser also for automation. And if you watch my previous videos, I created this framework or this setup called surf agent. So, we're going to use that too. So, these three things are basically all we need for like a super simple but very effective uh automation flow. So, I'm going to show you how this works today and how

I set this up. So now before we get into it, I just want to show you how uh I plan to do this today and what kind of automation we want to do. Okay. So for this example, uh what we're going to do is pretty straightforward. Uh we want to set up this Hermes AI agent use cases research. This is just an example, right? So we have this, yeah, you can call it these five steps here.

We want to do a search using the SER API. That is today's sponsor. I'm going to tell you more about them later. We do Google YouTube search. We search for yeah Hermes AI agent past seven days. We verify this using our surf agent by going to the browser factecking looking at all the sources. We're going to write a report. We're going to submit the report on a Google form I have created for this. So again the

surf agent comes in. It can do form filling on the browser in Chrome. So you will see that too. And then it's just going to clean up. So this is going to be the example we're going to run through today. So I think we're just going to head over to the Mac Mini and let's run this example so you can actually see it work in practice. So like I said everything here is going to be skill

driven and this recon skill is what is going to drive our automation here. Uh you can see basically it's a pretty standard setup because I want to show you how we are going to run this now. So this uh is the command that is going to be triggered if we put this into a chron job. I'm not going to do that today, but if you have a chron job, you can just trigger this command in bash

when that yeah chron job fires. So we do cloud-p and we do the slash recon in query. So if you know in cloud code this will actually trigger the skill. If I do like skill slash recon something like that this will of course trigger the skill we have created. So basically when we are running this this will actually trigger our skill. Uh I passed in the model here. I want to do cloud set for six on

this and I want to do dangerously script permissions because I want to allow all tool use on this example uh automated chron job setup. So basically that is going to run this skill right that is a chrondriven recon skill. So it's going to pull fresh SER API results from Google YouTube and news surf agents to fetch everything we talked about and basically that is the whole setup. So before I show you how this runs, now uh

I talked about SER API in this video, uh this is very crucial to exactly this setup. So yeah, let me tell you about today's sponsor, SER API. So we all know that context is crucial when it comes to building out your agents and skills. And we want this to be as reliable as possible. And today's sponsor, SER API, does exactly that. So I want to show you how I use SER API to set up my skills

in, for example, Cloud Code today. So, I just want to head over to their documentation here. So, let's head to my dashboard. And here you can see I find the documentation for the Google search engine results. So, I'm going to grab this and YouTube. We have that too. I think YouTube search API. This is also something I like to use. And I put all the documentation here in cursor. So, we have the documentation. Then I can

just head over to my cloud code. And let me do this prompt here. I'm just going to say read the doc SER API. Build a skill that uses the API key in ENV. Use the SER API to get contacts from Google search, Google News, and YouTube based on the user's input. And now we're going to build out the skills and let's test it and see how reliable this works now. Okay, so you can see Claude Cod

built the SER context skill here. So now, uh, let's test it out and see if we can do some quick research. Let's just do use the SER context skill. Search YouTube and Google for any news around GT 5.5 or SPUD. So this was released yesterday and this should bring up now all the latest freshest news from a simple Google search where we get this nice JSON output. I can show you see all the outputs comes in

this JSON format here and this is very useful for LLMs right so let's go back to cloud code now see what we have here yes you can see the launch we have all the info we gathered from the launch here officially launched yesterday April 23rd perfect what is the most viewed YouTube video on GPT 5.5 and you can see it's super fast it's reliable right and I can just say open it so this is why I

love using SER API when it comes to building out my agents and skills the reliability, the avoiding of captas, proxies, and everything. So, if you want to try SER API now, follow the link in the description or just scan the QR code on the screen. You will get 250 credits that you can play around with and set up your skills and agent. So, big thanks to SER API for sponsoring this part of the video. Now, let's

go back to the project. Okay, so now we want to run this. So, basically, the way I have set this up, you can do this uh a bunch of different ways. I just hardcode my topic into the skill here, but you could pass this in like a bash or something if you wanted to change it up. But today, we're gonna do research on the Hermes AI agent use case, and hopefully we're going to end up with

a Google form that is filled out fully autonomously by Sonet 46. So, we can look at what it did find here. So, what we want to do now is pretty straightforward. Yeah, we have no tabs open. So I'm going to head here and I'm just going to run this command now. And that is all we need to do, right? So I'm going to start this now. So of course this first part now using SER API is

headless. So we won't see anything. But since I'm running this on like a dedicated Mac Mini, uh I can do the the browser with my surf agent setup. Uh I can do that uh without headless. So this could be we can watch what happens. So first now of course it's going to do some SER API look up finding some YouTube videos and then we're going to go into the surf agent and you will probably see the

browser is being triggered here on the screen and Claude is looking at all the sources it found on the Hermes agent. So let's just wait for that and then we can follow along what happens. Okay, so now you can see the surf agent is launching the browser, right? It's looking up Hermes. It found a video here on a YouTuber talking about the Hermes agent. And remember, it's going to look for three videos. And you can see

on the right here, it's looking at the transcript. It's gathering all the information in transcripts. It's going to scroll down hopefully and look at some of the comments to get some information from there. Let's see if it's going to do that. Yeah, you can see it's scrolling down, looking at the comments. Now, it's probably going to do the next video and see what happens here. Yeah, here's the final one. This is a different uh video. And

again, probably going to check the transcript. Look at the comments. And now you can see it's heading over to Reddit too. And you can also see it's checking its sources on the web here. So here it is on decrypt and it's on another website, scrolling through the sources. So it's basically doing some factchecking on what it looked at. And now we go to search hacker news to see if there are any interesting stuff there. So, it's

basically doing like a mini recon, mini research setup, looking at the topic we had. And finally, it's going to end up on that Google form and fill it out. And we can access this from our mobile, right? Okay. So, we closed the tabs here. Uh hopefully now it's going to head to the the Google form and fill out that. Yeah, perfect. Here is my AI agent automation recon form. So, we have three questions. What was your

most interesting idea you found from the research today? Uh we're going to leave the URLs of the sources and it's going to rate the idea from one to five stars. Right? So that is all the forms we have and it's going to click submit and I can access this like I said from my yeah you can see here it puts in like an omnivariable device plus obsidian vault plus her agent creates an always on ambient memory

loop. Okay. It leaves the URL for a source of that ID the YouTube and it rated it five out of five stars and that was sent and that is the whole loop. Now if you head back here now I think when this ends we get like a summary in the in the terminal too of what happened and that kind of completes the loop and the chrome job stops and the next day you can run this again

and you can also have bunch of different chron jobs you wanted to set up with the different topics right so it's very flexible so let's see here now yeah we got this five novel ID Hermes uh yeah here we have some more ideas, but remember it picked the one that it thought was the best, but we ended up with total five ideas here. We can continue working on if you wanted to. Okay, so we give a

four out of five rating, not five. Okay, that was my bad. So, yeah, that is the full loop fully automated research plus factchecking using surf agent and stuff like that. So, let me show you the form now and let's see if our response got recorded. And you can see now we have one response. So, I'm going to click on this and yeah, you can see we have it. Yeah, we got a four out of five and

everything works here. So, this is a super simple way to kick off this uh automated uh setups and basically you can turn it into whatever you wanted. It doesn't really have to be this kind of research but with the simple framework of using like the cloud-p with the skill setup and you can add in like today's sponsor SER API that is also really good but then because you can search Google and you can search YouTube and

a bunch of other things too like I said in the the integration and of course my surf agent can do everything browser fill out forms and everything. So you'll find a link in the description also to my surf agent. So yeah, that is what I wanted to share today. I hope you enjoyed it and don't forget to check out SER API. You can find the link in the description and you will get 250 credits like I

said. So yeah, thank you for tuning in. Hope this gave you some inspiration of building out your AI agent automation setups and I will see you again soon.