NEW FREE OpenClaw Update - GPT 5.4 + Gemini 3.1 Flash-Lite + Agents

Today [clears throat] we have a brand new update from OpenClaw. It's a new free update. If you've been sleeping on OpenClaw, wake up. This new release just dropped and it's packed with features. We're talking about GPT 5. 4 support. Gemini 3. 1 Flash Light Slim Docker builds and way more. Let me break it all down for you so you can see what is actually inside the update and so you can actually use this stuff. Even if you've never heard of OpenClaw, let's get straight into it. If you're not sure what open core is, by the way, it's a free open- source AI agent platform. It's like a brain. You can connect to almost any app. You can set it up on Telegram, Discord, Slack, etc. Searches web, run code, uh manages files, uses tools, even spawn sub agents to handle tasks. It's like having your own personal AI assistant that lives inside all of your messaging apps at once. Now, why would you care about this update? Well, there's a bunch of new features and headlines. You can see GPT 5. 4 4 and Gemini 3. 1 flash light integration. So basically you can use the newest updates of GPT and Gemini. You also got ACP bindings uh slim docker secret ref pluggable context engines hif image support and zo channel fixes as well. So we're going to get straight into this. You can see the full change log right here. And there's a ton of new updates. It's absolutely packed with features. If you're wondering what it means, etc., um, what all this means for you and how you can actually use it. I'll talk you through each one in a second. If you're not sure how to update, literally all you do is you just go into the chat of open claw and you just say update for me, right? Um, and then you're good to go. And you can see, for example, here we got updated to the um 7th of March version, and I asked what API using and it's now running on GP 5. 4 codeex ath, which is awesome. Uh, so it's pretty easy to set up, pretty easy to update, pretty easy to um use. Now, let's talk about what it means for you. So, here's the seven biggest things you need to use. And basically, if you're not sure, basically, there's new updates that come out all the time for OpenCloud, right? And often they make them like faster or safer or more powerful. So, the first update that you can see is GPT 5. 4 and Gemini 3. 1 flash light support. And this is a headline feature. So, Open Claw now supports the latest AI models out the box. GPT 5. 4 is now the default model. When you choose OpenAI, that means if you just type in like GPT in your config, automatically uses GPT 5. 4. No extra setup needed. And then also the other one is Gemini 31 flash light. So this is Google's super fast, super cheap AI model. Perfect if you want quick answers without burning through API tokens. Openclaw handles all the weird ID model stuff behind the scenes. You just pick your model and you're good to go inside the on boarding. Now the next bit is pluggable context engines and this one is a bit more technical but stay with me I'll explain it as simply as I can. So when you chat with an AI it has a memory window right it can understand how much of the conversation it can see and when that window gets full open crawl has to compact aka shrink the conversation. Now before the update there was only one way to do this. Now you can actually build your own context engine plugins. That means you can swap in totally different strategies, memories, uh, for how your AI remembers things, right? It's actually a new plug-in called lossless claw that tries to remember and keep more of the conversation intact, right? So, it's better memory across your open claw. It's kind of like upgrading your AI's brain from like a notebook to a filing cabinet. So, if you're just a regular user, you don't need to do anything. Everything works exactly the same as before, unless you install a context engine plugin. Next up, ACP binding. So if you're not sure what ACP is, basically this is a way of like using agents inside Open Claw, you can see like the full details and configuration how this works with um OpenClaw normally. Now with ACP, basically this stands for agent communication protocol. It's how OpenClaw agents talk to each other and to specific chat channels. Now, before this update, if you restarted OpenClaw, your Discord channel and Telegram topic bindings would disappear. That mean your agents would forget which channels they supposed to be in. Now, with those bindings, they're saved permanently. So, if you restart the server, your agents pick up right where they left off. This also works with Telegram forum topics, and you can even bind a specific agent to a specific Telegram topic thread. And there are now approval buttons in Telegram so you can approve agent actions right before your chat as well. Next up, we have Slim Docker multi-stage rebuild. Now, if you're not sure what Docker is, basically it's a way of setting up a container for your OpenClaw, right? You can get it for free. It's just kind of like a safer way to host stuff. You can actually switch things on and off as well. So, if you set up OpenClaw, then you want to switch it off, you can press stop inside Docker and then you can add it back in later

right? That's how it works. And so with the slim do docker multi-stage rebuilds um if you run open core and docker which is one of the easiest ways to get it set up on the server the docker image is now built in multi-stage format. That's a fancy way of saying the final image is way smaller. All the build tools source code and extra stuff gets thrown away. You only keep what you actually need to run open core. And there's even a new slim option you can use. You can use a variant. It's called openclaw equals slim when you build. And you also get another feature called openclaw extensions. This lets you pre-install your favorite plugins right into the docker image. So when your container starts up, everything is already there. Basically faster startup, more reliable, and less headaches. Um there's some security upgrades as well. So before people would put like passwords and tokens and stuff directly into the config file, which isn't ideal. Now there's something supported called secret ref, right? So instead of putting your password inside the config, you put a reference to where the secret lives. Um and it's like pointing to an environment variable or a secret manager. This means um your passwords are not sitting in a plain text file which is more secure. Of course, we have H EIC image support as well. So basically if you take photos with your iPhone, unless you change your settings, they're typically going to be in a format called HIC, right? And most AI platforms can't read them. OpenClaw now automatically converts HEIC images to JPEG before sending them to your AI model. So if someone sends you an iPhone photo in Telegram or WhatsApp and it's HIC format, Open CL just handle it. Um which means you don't get like any unsupported format errors as well. And then there's a ton of like channel fixes across the board. So there's literally hundreds of fixes for every chat platform open supports. Some highlights. So for example, you got Telegram. Um, so streaming messages no longer show duplicate bubbles. Draft previews work properly in our DMs. Polls are supported as native action. Native commands work correctly. Media uploads. You can see a bunch of updates inside Discord as well. Slack, WhatsApp, Lime, Fu. Facu is a Chinese messaging app. You got Signal as well. And you got uh Matamos as well, right? And then there's a bunch of security updates. You're going to see this as a common theme across all the updates that come across inside Open Core. So if we scroll down, we can actually just control and f this and you can see a bunch of fixes um for security right here, which is obviously like a big concern with open call being an open-source um example, right? And then there's a bunch of uh memory and compaction improvements. So memory search with QMD, the inbuilt uh knowledge base got several fixes. So keyword search rankings now work correctly. Uh collection conflicts are handled automatically. Duplicate content errors trigger an automatic rebuild and retry. This is some technical stuff. Um, but basically what this means is that you can now choose which sections of your AI agent instructions get reinjected after compaction. And this gives you way more control over what your AI remembers. And then there's something called internationalization, right? So the web control um panel now supports Spanish as a language and it includes local detection, lazy loading, and language pick of labels. And I think there's a bunch more languages coming soon. I've been seeing them inside every single update, adding new updates, um, adding new languages, etc. So, just a recap on the whole update. Quick recap in 30 seconds. GPT 5. 4 is now the default OpenAI model. Gemini 3. 1 Flash Light is fully supported from Google. There's pluggable context engines that let plugins control how your AI manages memory. ACP bindings, persist through restarts, slim Docker builds, secret ref for storing passwords, HIC images, auto convert, hundreds of channel fixes, major security hardening, memory compaction, smarter and more reliable, Spanish language support. Um, and that's basically how it works. All right, so if you want to get started, literally all you do is you just go inside the chat and you say update to this. Sometimes you have to, for example, um, restart the gateway, but typically it's fine. And that's the whole update right there. So, thanks very much for watching. If you want to get all the video notes from today and you want to get a six-hour course on OpenClaw, plus another three-hour course right here, and daily updates on the new latest updates. So, for example, recently we compared Keyo Claw versus Open Claw. We also looked at Claude Code skills, which is a way of kind of making um claw code up to the same standard as OpenClaw. We also looked recently at some of the best free open core alternatives including 30 different versions. So if you want to get all of this and daily updates like you see you can get that inside my AI automation community the AI profit boardroom. This is a community focused on helping you save time grow and learn with AI. Right? There's 2,600 members inside here which means you can ask questions, you can get help, you can get support whenever you need to. You can also for example join our daily accountability groups. You can post your goals and stay laser focused. We do a

weekly AI update. So, we cut through all the noise and we just give you a filter on my 60 hours of research. And basically, I look at everything that I've researched, all the stuff I've tested. I go live for about um 90 minutes a day as well. And I just compact all of that information into a weekly AI update inside the AI profit boardroom so that you can save time and also you can see what other people working on as well, which is pretty awesome as well. Right. Inside the calendar, you can get weekly video coaching calls. You can get help and support whenever you want. Inside the map, you can meet people in your local area and connect with them, which is pretty awesome as well. And then inside the classroom here, you can see, for example, our best training. So, you can go from beginner to expert with AI automation and learn how to build your first AI agent in under 5 minutes. You can also, for example, get my best playbooks on how I automate X, newsletters, shorts, Instagram, um, AI avatar videos as well. And then if you want to get the daily new updates, the new upgrades, you want to find out what's working right now, etc., where we do daily video tutorials with step-by-step guides inside um this section right here. Every single day, we add multiple guides like you can see each day based on what's working and what's new, etc. Then, additionally, you can see how to get more agency clients with the agency course. You can watch the coaching calls back if you don't make them live. You can learn how to rank number one with AI SEO over here. And you can also learn how to grow a YouTube channel with AI based on what's working for me. So you can learn all of that inside the AI profit boardroom link in the comments description or go to the aiprofit boardroom. com. So today we're going to be testing out GPT 5. 4 with Open Core. If you've never done this before, um, basically there's a new update from Open Core that actually makes GPT 5. 4 the native model if you connect chat GPT to open core. I actually set this up this morning and the update just came out a few hours ago. You can see for example, we've asked what API you using and it says the session um is running on open AI codeex GPT 5. 4 via open claw, right? Um, this is a powerful way to use this. Now, this is really interesting because GPT 5. 4 is not just like another model upgrade. Um, it's one of the first AIs that can literally control your computer like a human. And if you're wondering, okay, like why should you care? Um, well, imagine your AI assistant could actually do things on your computer for you. Not just talk to you, not just give you answers, but actually open apps, click buttons, fill out spreadsheets, browse the web, or complete full tasks all on its own. That's basically what GT5. 4 can do. And when you combine it with open claw, you now have a personal AI system that works across your gateway, WhatsApp, Telegram, Slack, Discord, and 20 other platforms powered by one of the smartest AI models ever released. Now, the other thing as well, if you're using, for example, uh Claude, well, Claude can get pretty expensive, whereas for example, with this, you can go um directly here. Now, Oper says, "Do you have a channel in Spanish? The Latin community wants to learn from you. " Thank you very much. So you can actually uh after 24 hours of the videos being uploaded, they get automatically converted into languages like Spanish on YouTube. So you can watch them back like that is it's a bit annoying with the 24-hour thing, but I think YouTube just gets uh faster and it's a pretty cool way to learn for free. Now before we get into this quickly, what is open claw? So let's start from scratch. Openclaw is a free open source uh personal AI assistant you can install on your computer or server. Think of it like having your own private chat GPT but way more powerful and personalized, right? So it connects to all the messaging apps you use. You talk through those apps and it responds using AI. You can also browse a web control a browser run schedule C uh run scheduled tasks, manage files and use skills, right? And the key thing is you choose which AI brain powers it. Uh you could use claude, you could use GPT, Gemini, Grock, etc. or even local models running on your machine. And that's where GPT 5. 4 comes in. Now if you're wondering what is GPT 5. 4, GPT 5. 4 is OpenAI's newest and most powerful AR model. It was released on March 5th in chat GPT the API and codeex and it brings together the best of OpenAI's recent advances in reasoning coding and agent workflows into a single model. Now there's been a lot of um a lot of people saying that they didn't like for example GT5 and those sort of models but I would say GT5. 4 is a lot better than previous versions of GT5. And um if you're wondering okay what are some of the biggest upgrades? Well, I'll talk you through those in a second, but we'll just get straight into this. So, for example, we're going to do a test task here. So, I'm going to say, okay, um

let's see how responsive it is. So, I'm going to say, you know, what is 2 plus two? And we're just going to see how fast it is. Um, just to see if it's fast enough when it responds. Now, it's pretty quick when it responds like you see right here. If you're using something like Gemini 3. 1 Flashlight, that's going to be even faster to be full to be fair versus something like GPT 5. 4. for but at the same time it's quick enough to use and the other thing about uh using chat GBT codeex is that the limits reset every single week right and so I've never hit the usage limits on openclaw with GPT 54 any other GP models using the codeex a whereas for example if you're using something like u claude a well number one claud like cutting back on that and they say you shouldn't do that but number two is you can hit the weekly usage limit like really quickly uh within Claude and it also has session limit usages as well. So if you're using GPT 5. 4 typically you're going to get more out of it. tokens. You can use it more often. Um frequently without running into session limits. So that's a big benefit too. Now we're going to test browser control. So, I'm going to say, okay, open up a browser and check in the latest AI automation news. Let's see if we can do that. See how quick it takes as well. Um, and what I'll do is I'll share my whole screen so you can see what's going on in the background. So if we have a look here, it's actually opened up a browser and then we just have to allow permissions like you can see. And now it's beginning to look through permissions. It's using the browser. Um it there were actual benchmarks tested on this and actually found that with GP 5. 4 it can use a computer better than a human apparently according to the benchmarks. Now if you believe that benchmarks are not I'm not here to debate that. I usually don't pay much attention to them. Um but you can see how it works right here. And it's also got web fetch as well as a tool. So it's using that to come up with ideas. It's now finding a bunch of stuff as you can see. And it's researching the ideas right here and then it's replied really quickly. Um, so now it's come back with a bunch of ideas on the latest news and that sort of thing. Um, it seem to use the browser pretty quickly. It's very responsive as well, which is pretty good. does seem to be a lot faster than previous versions of GPT as well. I've used It's actually giving pretty good answers on the responses as well. So, for example, it's looked at the new um OpenAI codeex update, given us a full breakdown, formatted it pretty nicely, given us a stats breakdown as well. It seems to work really well natively with OpenClaw, like better than anything else I've seen previously. So, it's pretty good for that. Um now it says you know we can do a second test here as well if you wanted to find out information um even more information but basically it has got a bunch of new updates and these are new things that recently came out so it's pretty useful. Let's say uh schedule this task um to happen daily at 4 a. m. See if we can do that. And then it's given a suggestion on how we could prompt the scheduled update as well. And then we should have that scheduled in. Done. Nice and easy. All right. So, that was pretty simple and easy. Let's

try something a bit more complex like for example installing a GitHub um repository. So, what I'm going to do from here is we're going to install this workflow like you can see. And this basically turns um openclaw into like an agency, right? So an agency team. So for example, you can create agency agents and then they give you a full specialized team that can for example be like one can be a front-end developer, backend architect, mobile app developer, etc. Right? We can plug this into open cloth to just make it a lot more powerful. So I'm going to see how long this takes to install. and set the stopwatch over here and then we'll install this and see how long it takes and whether it can actually complete the task properly or not. So you can see here it's breaking down the installation and the GitHub details. And then it says it should be available Now to CL code to use on this machine. And that took 1 minute to install or 1 minute 15 seconds. Pretty cool. So, let's talk about the seven biggest upgrades inside GPT 5. 4. Number one is native computer use, right? So, GPT 5. 4 is OpenAI's first general purpose model with native computer use capabilities, enabling agents to operate computers and carry out complex workflows across applications. This basically means the AI can literally look at your screen, move the mouse, click buttons, type text, and navigate software just like a human would. Right? Before GPT 5. 4, if you wanted AI to do something on your computer, you had to build complicated scripts and workarounds. Now, it's built right into the model. Now, the interesting thing here is on the OS world benchmark, which measures how well AI navigates desktop computers, um, GP 5. 4 hits 75% whereas the human expert score is 72. 4%. 4%. So that's the first time any AI model has beaten humans at controlling a computer. Number two, he's got much smarter reasoning, right? So G5. 4's individual claims are 33% less likely to be false. And it full responses are 18% less likely to contain any errors. Right? So basically, it makes fewer mistakes. It hallucinates less. It follows your instructions more accurately. And this is huge for anyone using it inside OpenClaw because you need the AI to be reliable, not creative with the truth. Additionally, a big upgrade here is the context window. So, it supports up to 1 million tokens of context, allowing agents to plan, execute, and verify tasks along long horizons. What does that actually mean? Well, it can remember and work with roughly 750,000 words at once. That's about 10 full novels worth of text in a single conversation. So, for OpenClaw users, this means your assistant can handle much longer, more complex tasks without losing track of what you asked for. And then it uses way fewer tasks um and way fewer tokens, right? So, OpenAI reports that GP5. 4 uses um 47% fewer tokens on some tasks than its predecessors. Now, in terms of costs, this is not really a big issue unless you're using the API, but I thought I'd cover it. So, with the OAR, you don't really need to worry about this so far um because you're unlikely to hit the usage limits anyway. But basically fewer tokens equals lower cost equals faster responses, right? So if you're running open claw with an API key and paying per token, this is a direct cost saving. Um tool searches well. So previously system prompts would lay out definitions for all available tools when calling the model. The new system allows models to look up tool definitions as needed resulting in faster and cheaper requests. Right now open uses tons of tools like browser um schedule tasks, canvas, skills, etc. Before GPT 5. 4, before every tool had to be described in every single message eating up tokens. Now the AI only loads the tools as she needs for each task.

Now before we move on to the next updates, let's just test something else out now. So what I'm going to do over here is we're going to take this information like so. And this is for my AR profit boarding community. I'm going to see if open can build out something nice when it comes to creating a website. So I'm going to say create a beautiful modern interesting and fun to use website for my AI automation community. The AI profit blog and then open it up like so. We'll see how that performs. Let's see what questions we got here. Kobe says, "No walking pad. Not today. " Um, Riley says, "What's the odds we can run open claw locally using all three models for open router? " Um I haven't the only free model I used with open claw and open router was directly with uh something that was oh it was pony alpha and that was basically like a stealth model for the new GLM release. Um, but I would recommend if you want to run free models, just run it locally through uh, OALMA and you can run local free models like Quen 3. 5 or GLM 4. 7 Flash and those models will run uh, better, but you do need a good setup if you're going to run free models on open core. So, you need like a Mac Studio uh, minimum really to run like at a decent spec. So, it's going to build out a landing page right here as you can see. Let's just set the timer. We'll see how long that takes. And you can see it's just going to work in the background and start working through that. In the meantime, what we can do over here is just run through the last few outputs. So, um, the other thing here as well is like GPT5. 4 has its best ever professional work output. So, on OpenAI's internal benchmark for spreadsheet modeling tasks, GPT5. 4 scored 87. 3% compared with 68. 4% for GP5. 2. spreadsheets, presentations, financial models, um legal analysis, GPT 5. 4 outperforms most. Um and then you've got better vision and document reading. So on MMU Pro, GPT 5. 4 achieves an 81. 2% success rate without tool use. And on Omni. bench, average error rates dropped as well. So basically it can see images better, read documents more accurately, and can pro process full resolution images up to 10. 24 million pixels. So it's still building that out. As you can see, if you come back here, basically what's the before and after? Well, you know, before it could browse the web in a basic way. Um, it could answer questions and summarize things, but it struggled with long complex multi-step tasks, right? It would burn through tokens fast. Um, it couldn't reliably control your computer. Sometimes lost track of instructions over low long conversations and professional outputs were not that great. Now with OpenClaw plus GP 5. 4, you get faster responses. You can now reliably control your browser and desktop apps. It can complete multi-step workflows. It supports hot swapping, right? And then it handles 10 times more context, meaning longer, more complex SAS without breaking. I don't know if I wouldn't say that's 100% true. I would say it just handles more context, not 10 times more. Um, it uses nearly half the tokens for the same work and it's dramatically more accurate and less likely to hallucinate. [snorts] So, let's see how it's performing. So, it's still running through that as you can see. I'm going to just ask how it's getting on. It's been tuned itself up. So what are some real world use cases? Well, for example, you could build like a customer service bar. You could build automated research assistants, web automation agents, uh productivity assistants. So you could actually summarize Slack channels, answer questions from your documents, trigger automation tasks. Um and then finally, you know, desktop automation as well, which is a big one, too. And if you want to switch, just update your open claw first and then from there, you're going to just switch inside the onboarding to GPT 5. 4 and you're good to go.

So, it's still building out our website as you can see. Um, but just to recap on the key takeaways here. So, GC 5. 4 is OpenAI's most powerful model released on March the 5th. It's the first AI model with native computer use. It uses 47% fewer tokens on most tasks. It's 33% more accurate and 18% less likely to have errors in its responses. It supports up to 1 million tokens of context. and OpenClaw 3. 7 beta already supports GP 5. 4 natively, right? And you can switch with a simple command. And the best way to set this up really um is just use the O instead. So, thanks so much for watching. If you haven't already, feel free to check out the AR profit boardroom. This is my AI automation community that helps you save time, learn AI, and grow with AI automation. Inside the community, we have 2,600 members, which means there's always people online, so you can ask questions, get help, get support whenever you need to. You can also, for example, join our daily accountability group where you can post your goals and stay 100% laser focused on what you want to achieve today. Additionally, we do weekly updates where I basically type up what I've learned this week, what's actually useful, and what you can safely ignore. And then, additionally, you get weekly video coaching calls. You get help and support in real time on video calls. You can share your screen, ask questions, etc. On the map, you can meet people in your local area depending on where you are. So, you just zoom in and then you can DM people, ask to meet up, jump on a Zoom call, etc. And then inside the this section inside the classroom, you get all of my best trainings. So, for example, you can learn how to go from beginner to expert with AI automation and how to build your first AI agent in under five minutes. You can also, for example, learn how to automate X and newsletters and shorts and Instagram and AI avatar videos. And these are all systems that are working for me. And we basically give you the um full setup and the step by step guides as well. Now, you can see that the website has been built and it's um set up, it's working, it's running locally, etc. So, that was fully built with openclaw. It's even created like a nice little uh icon for the website as well. And this was coded with openclaw directly, right? to build out a whole website just using open core as you can see. So that took 5 minutes to build a whole new website using openclaw and we hosted it locally, built it out. Um it even previewed it for us and actually broke it down. It also said if you want I can do another polish uh to just make it more visual, more animated, more salesy, etc. Right. And then also you can get my daily updates plus a six-hour course on how to use openclaw and a three-hour course over here. And I do new daily updates on openclaw and all the latest things going on in AI automation here. Right? So this gets updated daily with new step-by-step guides, video tutorials, etc. So you never fall behind again. On top of that, you can learn how to get more clients with the agency course. You can watch about the video coaching calls over here. You can learn how to rank number one with AICO. And then you can also learn how to grow a YouTube channel with AI based on what's working for me. So that's all inside the AI profit boardroom. Feel free to check out link in the comments description or go to the AI profitaboom. com. Thanks for watching. Today we're going to be talking about how to make claw code 100 times better in one single click using a new extension called the free AI dream team, right? Um, basically what it does is it turns your core code into an agency um, using 61 specialists and is free to set up. So, I'm going to show you exactly how it works. It's a free open source project on GitHub that just exploded to over 12,000 stars. It could completely change how you use tools like, for example, Claude, Cursor, and Gemini. And I'm going to break down everything you need to know even if you've never touched GitHub before. All right. So, this is the update as you can see and basically what it does is it gives claw code a bunch of AI specialists that they can start using. Now, if you want to get started with this, you can just copy this command into claude code. So, let's get started with

that. Open up a new window here and we'll run claw code. We'll paste that in. And Quil Code can set this all up for you. Right. So, you can see we've literally installed that straight away. It says your agents are installed and organized into categories. Here's what you have. So you have an engineering team and a design team, right? Um plus you've got more agent specialists with marketing, product, product management, everything else, right? So these are basically skills that you install to your CL code or to cursor as well. And basically what it does is it gives you an AI agency um with like front-end developers. You've got for example um agents for all sorts of specialists as you can see, senior developers. There's a bunch of marketing agents as well inside there, which is pretty awesome. So, for example, like a content creator, Instagram curator, Reddit community builder, all sorts of cool stuff right here. Right. So, now what we can do is we can say, okay, use a front-end developer agent to build out a new website. So, let's test that out. I'm going to test this out for the AR profit boardroom and we'll use this new agency team to build out a new website uh for the AR profit boardroom community. Right. So, I'm going to say create a new website for the AI profit boardroom, my AI automation community using the content below and make sure you use the front end developer to build it. stick to HTML as a simple landing page. Copy below. And then I'll just add the copy from the website and we'll see if it can build out something better than we've already got. Now, what you could actually do is you could have like multiple um sessions open. So you could open up a new window here and you could have Claude running agents for something else, right? So we can say okay do this right and you can see how it's using the Reddit uh community builder skill that we've installed. So you can see it's using that front-end developer um sub agent and it's spawning one right there to build out the website that we just asked it to create. Now you might be asking okay what is the agency like what what's all this about you know the agency agents uh GitHub project well basically think of it like this right imagine walking into a building um and picking from like 61 different experts for example like a front-end developer a brand designer a Reddit marketing specialist right each one knows exactly what they're doing Each one has its own unique personality and style and each one gives you like real actionable deliverables, right? And so that's what the agency is. So it's a collection of 61 agent files you can plug into tools like cloud code, cursor, windsurf, gemini, CLI and more, right? And when you load one of these agents, your AI tool literally transforms into a specialist. So it's not like act like a developer. It's a deeply crafted personality with workflows, success metrics, um, code examples, and a specific communication style. Now, one thing I want to say before you install anything like this is I always check the skillmd files. Um, you just want to be careful before you install anything external. I do it for fun to show you as experiments on YouTube. Um, but if you're not sure about installing any of this or you don't want to do it, etc., don't install it. Like, just watch my videos and learn from it. And also check the skillmd files. You can run them through claude as well. You could even rewrite them yourself and then install them yourself instead. But the main thing is like just be careful when you do install skills to something like claw

code. So for example, when you activate the front-end developer agent, for example, your AI suddenly knows React um Angular deeply. It follows a specific workflow for building UIs. It measures um core web vitals. has a personality and voice that keeps things focused and it's like going from a random intern or something like claw code straight out of the box to a senior specialist instantly. Right now there's nine divisions inside the team that you install. So the agency is organized into nine departments just like a real company. So you got like the engineering division, design division, um and we've got the website ready to preview right here. So let's open this up. And you can also see uh inside the other window, we got the Reddit community builder um to give us like a 30-day quick start plan uh core principles and also uh tools that we can use as So, it's really interesting how it's built out. And um I say open this up. So, this is the new website we built. And boom, we've got the new website right there. Right. Um so, it's created the full website as you can see FAQs. Pretty nice setup. Um nice animation when it runs through the page. Super simple in the way that it's designed, but it looks nice, right? And so that was fully automated using that new agency team. Now that was just a quick example, but there's so much more you can do with this. And if we actually compare it side by side. So exactly the same content, but look at the way that it's designed, right? So this is the original. This is the new version, right? So you see how it's got like a nice little banner at the top. It's designed nicely. It's got like these nice UI elements. The new version looks me much better than the old version. This is the new version. Right? using that new um agency setup with the specialist. So, it's a pretty cool tool. You also got like a testing division support um specialized as well. And if you want to set it up, like I said, you can just go into claude code, give it that command. Um we've got the setup instructions inside the video notes from today in the AR profit boarding, but that's basically it. And then in terms of like real world examples, I've shown you how I built a new website using this agent team, but you could, for example, like activate five agents at once. Um and you could create like a you know a new marketing campaign right so you can say okay content creator develops a campaign um Twitter engager handles extract G etc right so let's try that so we're going to take this um example use case here we'll go back into the builder and we're going to say okay now create a marketing campaign for the AR profit border using the agency and then we're going to stack five marketing agents together. Right? So, we're going to create multiple sub agents together that are working together in cycl code. And again, like you don't need to be technical to use this. You're literally just speaking in plain language. It's really, really simple and easy to use. So it's now launching the content creator to develop the full campaign strategy and it's launching all five agents in parallel right now. So it set up the Twitter engager as well and you can see how it's working step by step. Right now you might be thinking as well at this point, okay, what makes this different from say just using your normal workflows? So, you know, if you're using generic prompts that say, for example, like act like a developer, the agency gives you a developer with a personality, um, a process, code examples, a specific workflow, right? And prompt libraries that give you like one-off prompts you use once and forget. The agency gives you like a full agent system you can use and reuse forever, right? So, um, it's a lot more specific. And if we just look at the numbers here, you can see how it works. Um, and it works with seven tools, right? So, it supports C code, cursor, wind surf, ADA, um, open code and add to gravity. I did try setting this up inside open claw as well, but with open claw, it seemed to struggle and it actually came back to me and said like it's not designed for this, right? So, don't use it inside openclaw. So, I think you could use it inside open claw, but it wouldn't be as

clean and it'd be a lot more messy than using something like claw code, which you can see like formats it perfectly um works in just a couple of minutes, sets everything up really nicely, and uh it's organized it nicely, too. So, you got the agent team here um all working together, the status, the deliverable, so what it's going to do. So, for example, like the Instagram creator will create 30 post concepts, um some scripts, story strategy, etc., right? Pretty cool. So [snorts] just to recap, the agency is a free collection of 61 specialist AI agents on GitHub. Each agent has a unique personality, workflow, deliverables, and metrics. It works with C code, cursor, winds surf, Gemini, CLI, ADA, open code, and anti-gravity. There are nine divisions. So you got engineering, design, marketing, product, project management, testing, support, spatial computing, and specialized. You can set it up literally in one click, as I've done today. It's open source as well, so you can use it however you want. It has 12,000 stars on GitHub. Um, I honestly don't pay too much attention to that because that can be manipulated. Um, but the main point is instead of using generic AI prompts, you can get specialists and you can set up an AI dream team like we've done today and create really good outputs like we did with this website. Um, that's basically how it works and the new website is much nicer than the old website. So, thanks so much for watching. If you want to get all the video notes from today, you can get that inside the AR profit boardroom. This is my AI automation community that helps you save time, scale, and grow with AI automation. It has 2,600 members. So, inside the community, you can ask questions, get help, get support whenever you need to. You can also jump on weekly video coaching calls. So, we have four weekly video coaching calls where you can jump on a live Zoom meeting, ask questions, get help and support in real time. Um, you can see our testimonials right here. So, we actually have over 135 pages of wins with people just getting awesome results with the AR profit boarding. as you can see. So, lots of cool wins here. And this is basically an infinite scroll of um you know, people getting results of this stuff, which is what we love to see. Also, you can meet cool people in your local area. You can join our daily accountability groups. You can post your goals and stay held accountable and see what other people are working on as well and connect with cool people working on the same stuff. Inside the classroom, too, you can also see our best training. So, for example, you can go from beginner to expert with AI automation and learn how to build your first AI agent in under five minutes. You can also, for example, get my best playbooks on how I automate X and newsletters and shorts and Instagram and AI avatar videos using this too. Additionally, you can get my latest SIP updates. So, basically what we do inside this section is we look at all the new updates, we create video tutorials on them, and we create step-by-step guides on exactly how to use them and implement them and what they mean, right? We also have a full six-hour course on open core here and another three-hour as well, right? And that gets updated daily and I talk you through every single update. Top of that, you can learn how to get more clients with the agency course. You can watch about the coaching course over here. You can learn how to rank number one with AICO and then you can also learn how to grow a YouTube channel based on what's working for me with AI. So that's all inside the AI profit boarding link in the comments description or just go to the AI profitboarding. com. Thanks for watching. Appreciate it. Bye-bye. Wise says, "Thank you. Happy to help. "