2026 Is the Year of Agent Orchestration! Enter Oz Agents by Warp

I think 2026 is going to be the year of agents and my favorite terminal browser which is Warp just released Oz. Oz is an is an agent orchestration framework in which you can create uh online agents in in the cloud that can do things asynchronously. They can do things well mostly asynchronously. um they can do things on a cron job as well. And when we talk about uh open claw and things like that which are basically on these cron jobs, it starts to open up this world where you don't need or you don't need to create AI agents anymore because a lot of these tools are going to have it built in. Now, this is the new warp which I think is overly simple. I don't know whether anyone remembers or or whatever, but there was a few more UI elements down here. What I'm going to try and do next week, if you uh bear with me, is I'm going to try and start creating some agents and having them work together. I was thinking like as I've been building with agents and I'm not doing any tutorials. I'm not like learning from like very little like one of the videos I watched was from your channel. I don't I haven't been doing a lot of tutorials. Mostly I've been just talking to the models and it's probably bad because it's the best if I would be learning from people like you. I noticed that I I I noticed this need that I wanted to create like a task list and apply multiple agents like um assign multiple agents to it and say tackle this but talk to each other like each one puts their name in front of the task so the other person the other person the other agent knows who's working on what and I guess this is kind of that but all set up for you. Yeah, exactly right. Uh, Oz, the orchestration platform for cloud agents. Oz is the orchestration platform cloud engine that powers all of Warp's intelligent features. Oz is designed to coordinate agents at scale, understand your code bases, executing tasks autonomously, and adapting to your workflows. Oz is multi multimodal by design, which is quite interesting because I think a lot something people want is this ability to be able to dynamically choose between different models depending on the task at hand. If it's coding, go and use anthropic. Whe it's helping you out with your car wash, go use chat GPT and thinking mode. You know, it's this idea of being able to um uh orchestrate the right agent the right time uh with the right model. So, you can run local agents and that's what I always do. I kind of you just type agent there and you can just write the request. It used to be that it auto detects. It still kind of does, but um they've changed that a little bit. Um O cloud agents run in the background on Warps infrastructure. So actually running a docker image up in the cloud. You just type cloud agent. Hit enter. And we can start to fill some of this in. So 0x5 website the repo here. Let's choose the repo docker image. We'll just click suggest. That's probably going to be node. Oh, what devase? And the sab command is going to be button install. And now I've created that environment. I can schedule my cron job weekly accessibility check of at 10:00 a.m. on the first day of the week. Give it my environment ID. Now I can give it a prompt. So check accessibility standards and make sure we are conforming. Now every week this will run and just make sure we're accessible. Do you use n? Yeah, I have. Yeah, I have used to react to Slack events, linear GitHub, custom web. Oh, interesting. schedules, recurrent tasks like dependencies or dead code removal. Now they they're definitely gearing towards code stuff here and I don't want you thinking that just because they're doing this um I think these are just easy examples but as an example I could because this is you you've got a machine you've got a computer and and for those more visual people when you see a terminal it's hard to think that this is this is my this is the computer you can do everything on your computer from this terminal that you can do with a UI a guey right Everything's the same. If anything, your apps, your browser here, when you press a button, it's triggering a command in the terminal. That's a really basic simplistic way of talking about this, but it's just to emphasize the power and importance of um of the terminal. You mentioned about MVM and and Webflow being slightly different than MVM, right? MVM is just an an application um node version manager um go help here. It's just it's just an application that doesn't have a UI to it. I have an N thing that takes our podcast, this podcast, when it goes live, I add in all the time stamps. I have an N function that I trigger to go and do that. Um that's it's to be honest that's great for like you know offline offline work or or you know if I'm in remote or something like that. But if I'm at a computer, if I'm whatever I can just I should be able to just trigger that with um clawed code. So, I built a skill to be able to do that and and whatever. Interesting. Saying that, I could have a schedule run. Again, I've got claw code installed on that remote machine because you can you can uh create your box and create your installation exactly as you want it. Um or just or just, you know, maybe not use claw code. Maybe I just use simple things cuz um I'm assuming uh Warp's infrastructure allows me to use MCPs that are installed on on that infrastructure as if I was in Warp uh to to basically run that task. You know, go get the go get the timestamps, create the LinkedIn posts that I put out on LinkedIn based on these discussions. All that to be said, I love Warp. This is not necessarily about Warp. I'm I'm really going to dig I'm going to start here. This is where where I'm going to start with with um agents and orchestration of agents. I think that's that's really where I'm getting at here. This is people are just catching up, right, to coding with AI. Theo we spoke about earlier, he's only really just getting into coding with AI. I'm thankful that I've been doing this for nearly a year now, and I'm already seeing that the next thing you guys want to start looking into is multi- aent orchestration. while everyone else is catching up with just single agent or maybe not single agent but maybe multiple agents but no orchestration between the two orchestration between the agents is I think where the next big learnings needs to happen so drop anything you're learning about one a well maybe maybe you need that as a backbone maybe you need that as a foundational learning but this is where you're going to need to put your effort in that's my prediction of 2026 guys so so heed my warning So, I'm going to start here. I'm going to start seeing how I can create some agents with a pre-existing platform. And then I want to start producing some tutorials on actually building the agents and building my own agents that orchestrate between each other because it's all well and good using someone else's um you know, infrastructure. How do I build my own agent that does things that are the way that I want it? Who knows? Anyway, maybe I'll I'll leave links to warp and everything down below so you can get started on it. It's It's amazing how this is just I think I get a,000 credits. I think you get a,000 credits for like your remote agent um for a limited time. So uh yeah, that's it. That's us cloud agent. This is part of a larger conversation on my show, Command AI, which we stream live every single week. We discuss the news and all things related to AI in the world of design and web. Catch us next week and join in the banter.