Google's NEW Cloud Managed Agents are INSANE!

So Google just released a new API for managed agents in the Gemini API and what that means essentially is you can have cloud managed AI agents building for you on a computer out in the cloud that is not local or running for you, but you can actually get it to go off and build stuff. I'm going to show you some of the cool stuff that we actually did with this. Now you can get for example like an anti-gravity agent working inside the cloud as a managed agent and there's all this coding documentation. Now what I actually did is I just built it into our agent operating system and I'll show you what it can do and how it works.

This is working with Gemini 3.5. So this is Gemini 3.5 working in the cloud as a bunch of managed agents. Obviously it's designed for agentic stuff and basically inside Google's whole ecosystem this is what they've implemented. Now if you're wondering where to get the Gemini API you can get it from AI Studio.

So this is AI managed agents with Gemini API and what this means essentially is like this is powered by anti-gravity and Gemini coding agents and you can use the API to build cool stuff or to do research. The way that I would actually use this is like if you just need a separate setup but you want to use AI agents you want them totally siloed away in the cloud and then also running 24-7 this is pretty cool and I'll show you some examples of how it worked for us. Now you can go inside AI Studio again I wanted to use the setup with the API so that I could plug it into my agent operating system and all you need is the API key from AI Studio and then you can go off and start using it and if you don't know how to use APIs or code or that sort of thing it's pretty chill because what you can do is you can just give it to something like Claude and Claude can figure out how to handle everything. So for example with Claude I actually just gave it access to my managed agents API and it plugged the whole system into my agent operating system and then I could see like okay what's it working on, what we're building, how is it going etc.

What have you done so far? So if we go over inside the agent operating system here we go to entity and then we go to managed agents you can see three things that were built. Now you might say well have you only built three things with this and I would say like it takes a long time because what these agents are doing is they're working for like 30 minutes at a time going off and doing stuff in the cloud and then you don't need to worry about it right so like you give it a task long horizon it works on that goal it gets stuff done and then it comes back later as you can see. Now the other thing that I would say that's kind of like a limitation with this setup is that yes it creates nice stuff as you can see right here for a beautiful research report but actually one of the biggest issues was like the UI on the first time around this is an improved version but before it didn't look that good when it gave me the research report so it looks good now and it's pretty cool what it's done and how it's done this whole research report and again Google managed agents as an API is awesome with anti-theft for doing research in the cloud but was it perfect first time around?

No. So like expect a little like back and forth is what I'm trying to tell you. You can see another example here so this is a research report on June and what happened in terms of all AI automation in June with AI agents and so what we've done is we got anti-gravity the Gemini API for managed agents and then we got these cloud managed agents to go off and go and do some research for us based on the latest stuff and then put it into a beautiful presentation like you can see right here and so this is called the June 2026 AI titans report an intelligence retrospective and it's looking at everything that's been announced so for example here I mean this has already been taken down that's how fast AI moves like I created this this morning and it's already been taken down but what you can see here is we've got Claude Fable 5 and Mythos 5 announced inside this report, Siri AI obviously that was a new announcement, Gemma 4 came out earlier this month it feels like even long ago that's how fast AI moves but it's only a few days ago Gemma as well and so it's actually really really good at research and then taking that research and plugging it into a beautiful report like you can see here and then we can flick through the presentation so I like the UI of this this this was pretty nice again we had to go back and forth on it a little bit but look how nice it is right and again the AI agents went and did deep research found everything compiled it into a report added dates then organized it beautifully with animations like you can see and if you asked for example Claude to do the same thing it wouldn't look as good what I mean by that for example is like when I normally get research on what happened in June for AI automation it's going to be super old boring stuff so let's try this for example if we go inside Claude compared to Google Managed Agents and we're like give me a research report on what happened in June for AI automation right so this is Claude and this is Google Cloud Managed Agents you see how beautiful this report is as well so it's going to start digging through some interesting stuff let's see what it finds and whether it orders it in in terms of the biggest news so if you look at this report for example obviously Claude Fable 5 and the release is the most important news that came out this month so that's why it's prioritized that first and then sort of batted the more boring stuff like for example you can see this note on like GP AI or for example M3 which is great but that came out last week and so this is going to give us a report now and let's see what it found so it's not really delved deep into stuff but we'll have a look in in a second and see what we get inside the report oh here we go it's already loaded so if we have a look here look how generic this reply is right where's where's the details on Claude Myrfos which is the most important thing that came out this month it's not there right where's the design it's not there where's the depth of the research and you could do deep research but again that would take ages whereas i do think like google cloud managed agents it's pretty cool for deep research because it's a long horizon task and it's just going off in the cloud and doing this and then we've got another one here which is basically looking at the state of AI coding agents we open this up in a new tab beautifully designed report we can click through it so we can see all the stuff that came out and again like KimmyK2.7 this just dropped yesterday and it's already found that news report and plugged it into our system so it's pretty cool how it's done there and how it's looked at all the agent profiles for AI coding and then broken that down into benchmarks as well so i think if you want some amazing reports you could plug antigravity into an agent operating system via the CLI give it access to managed agents and then those managed agents just create amazing research in the cloud and go off and build this stuff as well for you so that's what we're looking at and that's how powerful this stuff is and the interesting thing about this is you don't need infrastructure set up so you just give it a single API call with agent instructions tools and a remote linux environment so it's going off in a linux environment in the cloud and building this stuff and then it has persistent stateful environments so what that means essentially is like if you set up an environment in the cloud it's ready to come back to at any time so if i went back into for example antigravity CLI and then ask it to do other stuff it's going to be there ready to go and also antigravity is powering this now this is google's frontier level coding agent it's replacing gemini CLI in a couple of days and that's what's powering this whole system so it's pretty cool it's just like having a computer in the cloud but it's controlled by an API you don't need to sign up to a VPS or anything like that and then if you want a report on what it did in the cloud you get it back as code which is not that useful but you can always get for example antigravity to decode that for you and tell you exactly what to use so unless you're a developer i don't understand what it's saying there i mean it's give me the date and everything but i think it's better to just speak to antigravity CLI to control it because then you can kind of make it non-technical and you can still use the power of this stuff and so you get your own cloud computer it browses the live web it runs code and fixes itself the finished files come back to you inside your workspace which you've got over here which is pretty cool you could use this for research and it's an agent doing real work on a real computer this kind of like you know what this kind of makes it feel like is similar to perplexy computer but it's google out in the cloud right that's the difference here and so you can give it tasks you can go off and do stuff it's working in the cloud i call it the goldie cloud crew because you got a sandbox where you know you got this cloud computer doing every job in the cloud nothing touches your local setup you got the hands which is like it can browse the web it can go off and run code etc you can see a full log of what's been done from antigravity CLI if you want it and then you get the files back in your workspace inside your asian operating system which we've got here ready to go which is pretty useful and then it's on standby so once you set up the environment with the cloud it's ready to go whenever you need it as well which is pretty amazing and so like before you might i mean imagine if you're doing this research by yourself for example for a report like this back in the day you'd have to open 40 tabs run heavy scripts babysit the whole time you know people are scared of running ai on the local machine for stuff like this because it's a big task and then you have to sort of copy and paste the results yeah and you mess it's a lot messing around the new way is like you just give it one task it researches it runs code your mac doesn't have to do anything and if you don't have a powerful setup as well it's pretty good and then you get the the file back in your workspace as you can see right here with all these managed agent tasks so i really like it you can give it a super heavy job walk away so for example we actually gave it one keyword ai agent operating system and told it to do the seo research and it opened up like the actual search results read the pages and then create a full report brief as you can see right here on the whole setup which is pretty amazing so yeah that's basically how it works now in terms of what you learned you learned how to have a cloud setup via an api you got real data you've seen the finished jobs and also you stop losing the stuff that you do because you get it inside a workspace as well which is ready to go so if you want the whole setup that i've got right here with the agent operating system i mean you can use the antigravity cli directly with the gemini cli but if you want my setup where i've used this to organize it and then i've got hermes and i've got gemini plugged in i've got open claw code code etc this is all inside the agent operating system you can grab it inside the airpuff world link in the comments description or go to the airpuffworld.com and then you can turn claude you can turn hermes kimmy n2 and this whole cloud computer setup into one system with shared memory and one dashboard you control and you know you can actually have local agents running for free or you can have agents in the cloud with the managed agents api so if you want to get that it's all inside the airpuff world link in the comments description or go to the airpuffworld.com if you go inside the classroom and then you go to the new daily update section here you'll find the agent os with video tutorial the last update date so we just updated it today and then you can see the resources with the zip file in store and we add new daily tutorials it's a very active community based on what's working i just test the latest stuff and then show you how to use it inside the community you can jump inside the community ask questions get help and support in real time as well and then also inside the calendar you can jump on four weekly coaching calls where you can share your screen ask questions etc you can also meet people in your local city from the map who are using ai agents like you so you can easily connect with people who are doing similar things to you as you can see right here and that's all inside the AI Profit Boarding link in the comments description or go to the airpuffworld.com thanks for watching cheers