Claude Code for Business: Run Your Entire Company With AI Team

If you think code is just for developers, you're missing something huge. Clo code runs my entire business. And I'm not a developer. This is not about coding. It's about business leverage. In this course, I will show you every feature and mechanism in Clo code that matters for business. Not theory, not fluff. Real tasks, real results. Clo code solves every limitation of ChatGPT, Gemini, or even standard code. The memory problem, the context problem, the scattered word problem. Gone. I am Darren Viner and I coach business owners in the world's first AI community. The AAA by Liam Otley. For the last three years, I've spent thousands of hours building systems with AI agents for real business scenarios. Now, let's dive in.

Okay, let's start. Why are we stuck with ChatGPT and other AI tools? First problem is context window management issue. We are limited by the context window and the models. Whatever the platform, ChatGPT, Gemini, take whatever platform, they all have the same problem. The model loses performance as the context fills. There is loss of details, the model forgets important stuff, and at the end, hallucinates. Another big issue is the siloed sessions. You cannot share context across chat sessions or across projects or between custom GPTs in an easy manner. And it's a real problem when you're working on complex projects with different information scattered among different folders, different sub-projects, or different locations inside cloud. ChatGPT and so on. Really, that's a huge limit.

Then there is another problem, which is the self-orchestration problem. In ChatGPT, you can use teams of custom GPTs, but you have to call them manually using the "@" command. In Clo code, you cannot do this. In Gemini, you cannot use multiple gems in one chat session. And so, in all those tools, you have to orchestrate manually, either with a command like in ChatGPT or copy-pasting stuff between chat sessions, between projects, between gems, between different chat sessions. And so that's painful.

Another thing that is painful in these tools is the connection with external tools. Let's face it, you can connect ChatGPT with connectors or Clo code with connectors. But it's still very early and it's still basic. Like, for instance, the Google connector, the Google Drive connector, Google Sheet, and Google Docs connectors. You can't, you cannot even have your custom GPT or your Clo code project write Google Docs. That's crazy. So it's useless for me. And in addition to that, even if you use OpenAI schema with custom GPTs inside ChatGPT, it's clunky. Let's face it, it's techy. It's not straightforward. I really, it's, I hate it. And when you use it, it's really painful. You have to always authorize connection and so on. That's my god, that's so hard for nothing, right? So the capability to connect with external APIs is really limited.

Then you have all these other tools like N8N and Make.com, Zapier, and so on. All these automation tools. And at the end of the day, when you want to do automated workflows, you need to combine all these tools. It's really techy. Once again, it's brittle. Even if you don't pay any money, you will spend time. It's expensive to build, expensive to use, expensive to maintain, and it becomes obsolete very fast. Think about it, OpenAI has just released Agent Builder when I'm recording this video, which means tools like N8N or Make.com, most likely they can be threatened by that. Who knows if N8N will still be something in six months? Not sure. I'm not making any prediction here because I don't know, and anything can happen. But just to show you that these complex tools, you take time to learn them, and then you have to learn something else because now it's not, it's not the best way to do stuff any longer.

Once again, what we do here is not technical. We want to use simple AI tools that are evergreen. Let's face it, if you work with GPTs, it will last longer than working with specific AI tools. And here, what I will show you, working with Clo code, Clo code is definitely the future, and it is there to stay for a very simple reason. Anthropic is investing massively on upgrading this tool, maintaining this tool, making it better, adding features, adding capabilities. So it's just the beginning, which means it's the right time to jump on board.

And side note here, Clo code is initially it's a tool that is built for coders, but actually, you don't need to know any code to work with it. And in particular, with what I will show you, which means using Clo code inside Visual Studio Code and with the Clo code extensions, you will see it's, it's like an experience that is quite close to a chatbot. So you don't need any tech skills to use what I will show you. And the original thing that we do in CCG and that I show you in this course is that we are using Clo code not for code, but for business-related work, using AI agents, using processes, systems inside Clo code, leveraging the capabilities of Clo code, but for business purposes. And that's really super powerful and super easy, actually, once you have the right system, once you know how to use the Clo code capabilities in a business context, that's pretty straightforward.

So now, let's see how to install Clo code in few seconds. It's very easy to install Clo code. I recommend to use Visual Studio Code. It's free. And then to install the Clo code extension. And then the only thing that remains to be done is authenticate with your Clo code subscription. And once again, I recommend you to use your paid Clo code subscription rather than the API because you can also use Clo code with the Anthropic API. But if you use the Anthropic API, you will pay for every token you use. And I don't like this, uh, this approach. Why? Because basically, you're paying for mistakes that AI will do. When AI consumes too much content, it's because AI is not effective enough to do the job in a limited amount of tokens. And so I don't want to pay attention to the number of tokens that the AI is using, right? So that's why I recommend to use a paid Clo code subscription.

So let me show you how it looks. First, you go in Visual Studio Code, and here you choose the right one for you. I'm on Windows, so I would download the file here and I would just execute it and I would just follow instructions. It's really straightforward. Then, when you are in Visual Studio Code, you go in Extensions here. You look for Clo code, and here you have Clo code for VS Code. You click on that, and then you can install. For me, it's already installed, so I won't do it. But you click install, and it will do everything alone.

And the first time that you launch Clo code, and to launch Clo code, you just have to be in a folder. So let's open a folder like this one, for instance. Once you're there, you just click on a file or somewhere, and you will see that here you can launch with the upper right corner. You can launch as many Clo code sessions as you want. And so the first time you connect with Clo code, before having this beautiful interface, you will have something that says, "Hey, authenticate." And then you can choose the method between API or between your Clo code subscription. And I recommend to use the Clo code subscription. And it's pretty straightforward. You click on it, it will open the browser, and then you just say "accept" or you just read what is on screen, follow the process. It's really, really easy, straightforward, and you're good to go. As easy as that. And you see, no coding, no nothing. It's pretty simple. And you will see Visual Studio Code is really simple because it's a bit like Windows. You have folders with files, and you can navigate your computer there. And then you have a bunch of features in Visual Studio Code. But that's not the point of this course to explain that to you. You will discover, you will see, it's pretty straightforward.

Just a quick tip. You can see in a Clo code session here, I have access to the past conversation. Here, I don't have any. But in a folder I have worked on, for instance, that one here. If I open another Clo code session now, I have past conversations. Okay. So that means I can find my past conversation if I need to. So, pretty simple. That's it. You're ready. First thing to understand is the memory system. This is what we will cover next.

Yeah. When you first connect with Clo code, or sometimes when your access token has expired, Clo code will tell you that you need to log in again. And so you use the command `/login`, and then you will end up on this screen. And then that's where you can choose between Clo AI subscription or Anthropic Console, which is the API, or this. I haven't, I've never used that. And so you click Clo AI, then you have this box. You say yes. Then it opens this in your browser, and then you just click authorize, and you're good to go. Then you go back to your interface, Clo code interface.

The `.clodevrc` file, that's your business brain, or that's the brain in a more general context. That's the brain of Clo code. The `.clodevrc` file is a bit like a system prompt. And actually, you can have a `.clodevrc` file in any folder you are working on with Clo code. This is some sort of permanent instruction file that you can use at in each folder and for each project you're working on. Really, it's like a system or if you have worked with ChatGPT projects, it's like the custom instructions for ChatGPT projects. Or if you have worked with custom GPTs, it's like the system prompt. And same for Clo code projects. If you work with Clo code projects, you have custom instructions for Clo code projects. So here, the `.clodevrc` is the same concept. The thing to understand is that you, for each folder, you can have a `.clodevrc` file. And so if you have a folder with subfolders and so on, you can basically have a hierarchy of project folders, and for each of them, you can have a `.clodevrc` file. And the cool stuff is that each time you open a Clo code session inside a folder, it will look for a `.clodevrc` file to get instructions and context of what we were working on in this folder or project. In this course, I will call folder/project, it's the same. I will use the same two words for the same concept. So obviously, the advantages of that is that you don't have to re-explain a project context or business context. You can have different instructions per project or per folders. You can compound the knowledge because each time you're working on a project, you can ask to update the `.clodevrc` file so that the permanent memory or the system prompt, if you prefer, is always up to date. So it's super powerful. You can also use that to have some mechanisms that Clo code will execute when it opens. You can, really, the sky is the limit with this super effect. Clo code can help you design the `.clodevrc` file, of course, but you can also decide and tell Clo code to put stuff you find important for a given project or given folder inside a `.clodevrc` file.

So let's have a look at the `.clodevrc` file. Here I am in my CCG business operation. And if I open a, uh, a project, for instance, this one, you can see here I have a `.clodevrc` file. Okay. And you can see here, this is a markdown file with information about the project, the scope, etc. Okay. And if I go, if I go in another sub-project, for instance, this one, once again, I have a `.clodevrc` file somewhere.

So now Clo code has memory and context with the `.clodevrc` file. What if you need multiple AI teams working in parallel? That's where sub-agents come in. Sub-agents are AI workers inside Clo code. The cool stuff with them is that they are specialized AI assistants. They have isolated context windows. Each context window has 200,000 tokens available to the date of this video. That means you don't pollute the main context when you're working with agents. The cool stuff with agents is that Clo code can call them, automatically invoke them, and Clo code can run them in parallel or can run them sequentially according to the work they are doing. The sub-agents have access to the files in the folder you're working on, and so they can communicate with each other using the files they have access to, or Clo code can also pass context. I've seen that Clo code can pass information from one agent to the other, sharing context. Okay, this is technical, it doesn't really matter. Just Clo code can orchestrate automatically sub-agents, sharing context between the different agents, even if each sub-agent has its own context window. So the advantage is you can run multiple tasks at once, so you can go faster. You can specialize them as well. So that can be a team of specialized sub-agents that work on the workflow orchestrated by Clo code, not by you. You don't have to do it. Of course, if you want to specifically ask Clo code to work with a given sub-agent, you can do it. No problem. But that's super useful. And at the end, what a sub-agent is, it's a, it's a markdown file.

So, let's look at this. This is my GPT Express Mastery team. It's my, I call that my HR department. This is where I build all my sub-agents in Clo code, and also where I build all my custom GPTs when I'm working on ChatGPT. Okay. And so an agent is, you can see here, in my HR team, I have several agents. They correspond to my GPT Express Mastery framework, if you've watched the course inside CCG community. And if I open one, this one, for instance, you can see that an agent has always the same metadata here. And in the metadata, I have the name, I have the description of the sub-agent, its mission. That's what Clo code will use to know what the sub-agent does, and also to invoke automatically the sub-agent if the task matches the description. You can specify the, you can specify a color, especially if you're working on terminal, so you can see what agent is working. And you can also specify the tool it has access to. It's optional, but you can do it.

The classical procedure to create an agent in Clo code is using the command `/agents` and here `configure agents`. I will have to continue in terminal because here I'm in the extension for Visual Studio Code. But if you want to make the `/agents` command work, you need to go in the terminal. So if I go in the terminal, it looks like this. And then I have the list of agents I have access to in my system. And if I want to create a new agent, I can just click here. And then I can choose between project or personal. If I put my agent in project, that means the agent is available at the project level in the folder I'm working on. If I put it in personal here, that means the agent will be available everywhere. That's a hierarchy of availability. That is true for agents, for skills, for MCPs, for all the key features of Clo code. You can either have them available globally or have them available locally. I won't create any agents because, once again, I don't use this approach to create agents. What I use to create agents is my, is my HR team. And for instance, here, here I have my team working on an agent. It's an example that I started just to show you. I just said, "Hey, I need an agent that can answer questions about Clo code for business." Then what did Clo code? It refined my demand, or at least it identified inputs it needed and started answering the inputs automatically so I can review. I said, "Okay." And then what it did, it created a folder that is this GPT project Clo business expert. It's in there, Clo business expert. And then it called multiple agents to work on this sub-agent creation or this custom GPT creation. First, the purpose, the purpose refiner, that refined the purpose statement. And here it's a better definition of what the agent is meant to do, better than what I did initially. And then you can see here that the todos of Clo code using different sub-agents to work one after the other, executing my GPT Express Mastery framework to craft custom GPTs. And then you can see a good example of parallel execution. So it's not sub-agents that are executing in parallel here. It's the web search tool that is used by Clo code in parallel. So that means Clo code is making these different queries in parallel to go faster. Okay. A bunch of queries to collect knowledge, and then it's fetching several URLs to collect the knowledge, some other web search queries, and it's still making some research. Okay. And then after, we craft the knowledge base, and then the system prompt, and then potentially some tools, and then I'm good. I have my custom GPT or my sub-agent ready to rock and roll. I call that my HR team. And this is something that is available in CCG here in the Clo code OS, so that you can plug it immediately, just copy-paste it, and have it available inside your Clo code environment without spending time setting this up. It's already done and available in Clo code OS.

So sub-agents handle complex tasks, but what if you need reusable workflows or standard SOPs, standard operating procedures that activate automatically and that Clo code can call in anywhere globally? Which means these are standard ways of working that you can use in any project, anywhere in Clo code. That's skills. Okay. Consider skills as reusable workflows or standard operating procedures. Which means when you know how to do something well, and you want it to be done always the same way by Clo code, you will create a skill. And these skills can be called in any folder, in any project. So the key feature of skills is that Clo code can trigger them automatically, can use the skill automatically according to the work you ask it to do. It will detect context match. Then skills are really optimized for context window using a mechanism that is called progressive disclosure, which means when Clo code uses skills, it won't load the entire skill. It will just start discovering what the skill is about, reading the beginning of it, and then reading more if needed. And that's really super useful for context window optimization.

Another cool thing with skills is that skills mix deterministic code. You can put scripts inside skills, and also knowledge files that can be processed by LLMs. Which means you combine both worlds: code, deterministic tasks, and LLM for flexible AI automation or AI execution, let's call it. And finally, another stuff that is cool with skills is they are sharable. Which mean, first and foremost, you can use them globally in your on your system, in many different projects, don't have to specify again, they are available. And also, you can package them and share them with other people, for instance, with your customers, or for instance, with your teammates. Which mean you can enforce SOPs, you can ensure that your teammates using Clo code, or your employees using Clo code, will use the right SOPs because you share the skills with them. And guess what? In the Clo code OS, I'm sharing the skills I use for myself and ideas for myself. I share them with you so that you can use them right away. It's copy-pasting, so simple.

So like I told you, the advantage of skills is context window friendliness. And you know how much context window friendliness matters for large language model performance. They're reusable across projects, globally, anywhere from Clo code, except if you put the skill in one single folder, then they are available only on that folder. But if you put them globally, and this is what I advise you to do, if you put them in the `.clodev` subfolder in your user folder, they are available globally. Clo code decides automatically how to use them. If you want to be sure that Clo code uses the right skill, you can also tell it to use the right skill. And the cool stuff is that you build once a skill, and you can use it anywhere. And guess what? You can also improve the skill, right?

So let me show you what a skill looks like. Okay, here I'm in the `.clodev` folder in my user folder on Windows. And so that means everything that is there is globally available everywhere in my Clo code environment. Let's check skills factory. Guess what? When Anthropic released skills a few weeks ago, I guess three weeks ago now, the first thing I crafted is a skill to create other skills. Obvious custom GPT, I created, or maybe the third one I created is a custom GPT to create other custom GPTs. And back in the day, the first sophisticated prompt I created, it was what I call, I used to call meta prompt, was a prompt to create other prompts. It was in February 2023. Anyways, so this is how a skill looks like. A skill has always a `skill.md` file. Any skill has a `skill.md` file. You can see here, there is another one, etc. And so a `skill.md` file basically has, once again, metadata. You will see that in Clo code, all files, all system files, let's call them that way, as always a metadata that help Clo code understand what it is, what it does, when to invoke it, right? So here, same for skills, we have `skill.md` identified with this metadata, the name of the skill, the description of the skill, as simple as that. And then you have a description of the skill, what it does, the process inside. Basically, it's defined how Clo code should do the work when it's applying these skills. You can see it's a pretty short file, right? And if we check the outline of it, you can see in this specific one, we, if we check the outline of the `skill.md` file for the skills factory, there is the metadata, there is a more extended description of what it does. There is a skill creation process. There is some troubleshooting. There is some reference documentation. And that's the progressive disclosure. When Clo code sees reference documentation, it is how I have some other documents. I have access to, so potentially I can retrieve them. Some example skills, some scripts, and some version history. Okay. See it as a table of content. And then in the skill, I have some other documents. So that's a knowledge about how to do the skill properly. And you can have also code. Here I have some scripts. So these are Python scripts that Clo code will execute when it's using the script. I can give you an example about how skills are used. For instance, I have a skill that is, um, my standard operating procedure to create new projects inside my business OS. That's the project creator skill. And so let me show you how it looks like in real life. If I am in my CCG business operation system. So that's the one. Let's create a project using a skill. Okay. By the way, this operating system has many super useful processes, mechanisms, and so on, that makes it super effective to run any aspect of my business. And guess what? It's available in Clo code OS so that you can copy-paste it with all the mechanisms and the routines already set, and you can tweak them and make them specific for your own business. And of course, the implementation calls inside Clo code OS are meant to help you customize it for your business. These implementation calls are group calls. We do about each system I share inside Clo code, and it's also Q&A code so that when you have questions about implementing your Clo code OS systems, I can help you back.

In the main topic, so here, let me illustrate how to use skills, or a skill is used by Clo code. So I will create a mock project so that, so that I can show you how Clo code project creation skill to create the project the right way inside my CCG business operation. So that will be a mock project, just to show you it works. "Hey, use the right skill to create a project inside CCG business operation. That's a mock project, just to demonstrate how you invoke skills to do stuff, specifically here to create a mock project, right?" Oh, that's a clunky ask I did here, but it should work. So here you can see it's calling the skill "project creator" and it's reading, it's discovering, it's discovering the skill. That's progressive disclosure. Then the process is normally asking me some questions about the project, but here it has understood it's a mock demonstration, so it's just answering itself, following a procedure. And then I say, "Yes, that's good, let's go." And it will create it. And now it will follow the process to create the mock project as per, as defined in the skill. Okay. So it will create project structure, generate the `.clodevrc` file, seed file, what you should have in any project, a README as well, then do some other stuff that are defined in the skill. Okay. I will stop there. Okay, if you have created anything, please clean it. It was just an example.

Thing to remember about skills: the progressive disclosure mechanism, the mix of deterministic script code, which is really reliable, and LLM, that is more flexible, and you can mix that inside a skill. If you want to ensure that the skill is used, normally Clo code auto-invokes the skills, but sometimes it might just try to do it on its own, especially if it's a simple task you're asking. So if you really want to enforce a skill, mention the skill when you ask Clo code to do something.

So definitely, skills are super powerful individually. But what if you run multiple skills and sub-agents at once? Call that parallel execution. Let's have a look. We already saw a bit how to do this with sub-agents or tool calls, like when we, I showed web search in the previous example. Let's talk a bit more about parallel execution. So the cool stuff in Clo code is that you can execute in parallel either tools, either sub-agents, and sometimes just code sessions. Okay. So the basic two approaches are multiple sub-agents in one session. And in this case, Clo code will invoke multiple sub-agents simultaneously to do a workflow. Once again, each agent will get isolated context windows, but can share information. Either Clo code gives different information to the different sub-agents to coordinate them, or you can use also files inside your folder that these agents have access to to refresh their context, share information, share their work. Right? The cool stuff is that Clo code orchestrates automatically. You cannot do this in ChatGPT, Clo code, or Gemini, or these other AI tools. Clo code here allows you to do it.

The other approach is to use multiple Clo code chat sessions, which mean you open several Clo code windows inside the same folder, or sometimes in different folders, and each session works on different topics. Here, it's you who orchestrate and check what they do. Something to understand: if you want to orchestrate multiple agents, or if you want to have Clo code orchestrating multiple agents, you can do it in parallel if there are no dependencies. If task A done by agent A doesn't need task B's output, for instance, and if they're working on different files. Okay, you want to avoid having two agents working on the same files at the same time in parallel because it's messy. So the advantage here obviously is massive time saving with parallel execution. And really, the way I see it is I'm the CEO, and I distribute work among my teams. And so that's why most of the time I'm working with multiple departments. Here, I'm working with my operational team in my business OS. Here, I'm working with my HR team, GPT Express Mastery. And here, I'm working with my content team. Okay. It's like I'm a CEO with multiple teams, multiple departments. I can distribute work, and they can all work in parallel. Okay, I can run some stuff with my operational team. Then, in parallel, I can run stuff with my HR team. And then, in parallel, I can work on some stuff with my content team. Now, if I check in my content team, actually, we have several sessions. So there, my content team is working on several stuff in parallel. Same here, if I check my operational team, I have several sessions in parallel. When I work like this, I'm really the CEO with multiple teams, multiple departments. I distribute work, they do their stuff in parallel. I check what they're doing, I give them feedback, I give them direction, and that's the way we can really have super high leverage in what we do.

I want to insist on how powerful it is to have self-orchestrated sub-agents by Clo code and the capability of Clo code to invoke tools, to invoke sub-agents, to invoke skills, and for tools and sub-agents to invoke them in parallel. So far, we've seen different capabilities that Clo code has, but what if you need to connect Clo code with external services, for instance, APIs? That's the next one. The cool stuff with Clo code is that you can connect your environment with external tools. Okay, so far in the course, we've covered internal tools, internal features that Clo code has. Now, let's talk about external tools. So first, APIs. Don't be afraid. That's really simple. An API is just an interface that allows software to discuss with each other. And usually, what you do if you have an external tool you want to work with, you check if they have an API, and then you check the documentation of this API, you put it inside Clo code, you explain what you want to do, and Clo code will figure out how to discuss with this external tool because, once again, an API is just an interface that allows programs to discuss together. And in an API, you have endpoints, which are internet addresses. And when you want to work with an API, you just find the right endpoints, the right internet address you want to use, and then you send authentication. You send a formatted message to this address, and then the address, the tool, external tool, does its stuff, and then gives you back an answer, right? And that's as simple as that. So typical use case is when you want to connect your Clo code with email, like Gmail, Google Docs, or to your CRM to add data or pull data from your CRM, or post on social media. Every social media platform has an API, which is, once again, an interface that you can discuss programmatically with. Or if you want to fetch data from a database, you will use the API of the database tool. Okay, you want to work with an external tool, find the API you need.

And if I go, for instance, if I take the example of Gamma, which is the API I'm using to create presentations automatically. So you can see in my business OS, I have some files related to the Gamma API. Let's suppose I want to connect to Notion. Okay, never done that before, but I could ask Clo code to, "Hey, find the documentation of the Notion API so that I can create notes in my Notion account, and I can pull data from my Notion account." Okay, as simple as this. And then Clo code will just make some web research to find the API documentation. Hopefully, it won't be too long. Search, fetch it, so read it. And then, once again, remember authentication, then sending formatted message, then receiving answers, and doing your stuff with the answer. So here, basically, what I would need to do is to get my API key from Notion, put it in a `.env` file that is here, where I put all my credentials. And then Clo code will write the script according to what I need to do. Of course, I can make that evolve. And if I just explain in plain English what I want, Clo code, knowing the Notion API documentation, can just translate what I want into, "Okay, what are the URLs in the Notion API we need to work with? What's the format of the data? What's the protocol to authenticate?" And then it will manage that and create the script that allows me to do this. Of course, I would test that and so on. But that's as simple as that. According to the API and how well they are built, and it depends on the external tool. You may take some time to make it work. But basically, with Clo code, you don't need any coding skills. Which means, once you know an API you want to work with, once you get API keys or whatever is necessary to authenticate with this API, you're good to go. You can create a bridge, and then you can get this external tool working with your environment. Okay.

If you see here, it's looking for the right format to discuss with the Notion API. Okay, I will stop there because actually I'm not using Notion. For me, Notion is a distraction. But I know that many people like Notion. So that would be a simple manner to connect Notion to your Clo code environment. Okay. You can see here, authentication, the API basics here. And then it's here, it's documenting how to create a page or notes in Notion programmatically. We don't care about this. That's the format, okay, of the message we should post, which means send to the Notion API URL to create a page. Okay. That's the type of format that we need to send. An example, if you want to retrieve something from Notion, we send that to the right endpoint, the right URL, which is database. Can you see here, the URL was a page, here it's database. And then, of course, a database ID, which is the identification of the database you want to pull information from. And that's it. And so now, here, "Would you like me to create a complete Python or JavaScript script for a specific use case?" I could say yes. I will stop there because, once again, I'm not using Notion. But just to show you that you don't need to be a coder, and the concept is pretty simple. API, once again, it's an external interface you can use to programmatically connect with an external tool. And that is just URLs, and you send messages to this URL, you get answers. Okay, you send messages which are requests on the URL, the external tool does its stuff based on the message you send. The only thing is that the message needs to be properly formatted, the authentication must be properly handled, most of the time with API keys. And then, once the tool has done its job, it gives you back the whatever results you needed, and then you can process it inside your Clo code. Very simple. Okay.

Advantages. Ah, we can stop copy-pasting between tools. And look, we don't need AI workflows. We don't need Zapier. We don't need Make.com. We don't need all that any longer. Clo code can do it. Like, side note, the only use case when you need Make.com or a Zapier or an N8N is when you want an automation to be triggered whenever you're not in front of your computer and when you're not working. If you want background automation, yes, these scenarios are useful. Why? Because Clo code works only when your computer is alive and when you're working on your computer. But in my case, I have very few use cases where I need these workflow automations. And most of the time, if I need them, it's super simple anyways. So when you have an external tool that you're using, being Notion, Figma, Miro, whatever, your CRM, Airtable, and so on, they all have APIs and they all have API documentation. And so you can make research with Clo code to find the API documentation and make Clo code ingest the API documentation and then create whatever bridge you want with these external tools. Okay, that works great.

This being said, there is a new protocol that makes working with external tools theoretically easier, and these are MCPs. It's worth having a look at this because it's a very promising standard and might probably be the best way to connect with external tools in the future. It's still in development, so it's not perfect, but let's have a look at this. MCPs are a standardized manner for AI agents to communicate with external tools. It's a protocol that helps your agent understand what are the tools available in an external app and call this tool properly. In MCPs, you can have also additional stuff like you can have prompts and stuff like this. But hey, let's keep it simple. So think MCPs as power adapters for different countries, different tools. Clo code is the device, the services are the countries or the tools, and the MCPs is the standardized adapter. Okay, so that's an easy manner to connect Clo code, Clo code, or any AI agent to external tools. So there is lots of hype around MCP because MCPs bring official plugins, pre-made toolboxes for a given external app that your agent immediately understands and can tap into and use. Right? That's the theory. Everything being standardized means it's easy to develop MCPs, and also it's easy for any AI agent or Clo code to use an MCP. It knows the standard, it knows how to identify the tools that are available, it knows how to use the tools and then apply the tools to the external app, right? And normally, it's auto-integrated, which means Clo code is able to understand what tools it needs, look for the right tools, then use it, and that's it. Okay. So you don't even need normally to request for the use of the tool, even if it's always better to enforce it to be sure that everything is done fine. Okay.

But the reality is that it's a new technology. So there's big hype, but it still needs maturity. I've seen many MCPs popped up with big claims, and actually, they are crap. Why? Because they are poorly developed, they are not stable, or you can install them in an easy manner on some AI agent like Clo code, but not necessarily include code like that's a mess. And even when you use them, actually, they're not necessarily fulfilling the promises. So if I were you, I would only use official MCPs that are supported by the community and that are well-documented and up-to-date, obviously. For reliability issues, I will be very cautious with open-source MCPs. Why? Because you don't know the quality of them. You don't know if they are up-to-date. You don't know if they're well-documented. That's really tricky. Theoretically, you can create your own custom MCPs because MCPs is just a protocol that ends up making it easier to connect your Clo code or your AI agents with external tools. Easier than having to do it manually, understanding the APIs and creating scripts to make API calls. But at the end of the day, this is packaged API calls, right? It just helps AI agents to automatically understand what are the tools so that they can invoke them automatically without you doing anything and understand immediately what are the set of tools endpoints that are available and how to use them. Okay. So it's, it's, it's really a protocol to make life easier for AI agents and leverage AI to use external tools properly according to context. That's, that's at the end of the day, it's a packaging. This is what it is, it's a communication protocol.

So when to use MCP? Use MCP when you have a proven official, ideally MCP for an app, an external tool you want to plug on your Clo code or AI agent system. If no MCP exists, I advise you, or if you can just find some crappy open-source MCPs, I advise you to use API scripts. The thing we've seen just before. Same if you need custom logic or specific use of external tool, use API scripts. Uh, let me show you how to install MCP and what it looks like. Okay. So, I like this, uh, this GitHub repo, Model Context Protocol, which is what MCP means. `/servers`. Why? Because it's an official one. You have lots of MCPs here. Let's pick one. We'll pick Playwright, which is a Microsoft MCP. It's a browser automation MCP that uses Playwright to run tests, navigate pages, capture screenshots, scrape content, automate web interaction reliably. So pretty much giving your Clo code, if you're using Clo code, access to the internet and visualizing web pages and working with web pages. So that's pretty cool, right? So if I click there, I go in this page here, and here I have the information about the MCPs. But since we are not coders, we won't check that. But just here, quick review, here, 23k stars, that means lots of people have used that and rated it. Here you have the releases, like the latest one is this one. It was released yesterday, which means it's an MCP that is actively worked on and updated, right? Many contributors, etc. It's a really good indication that the MCP is actually legit. And so if you want to install your MCP in Clo code, you can either ask Clo code directly. "Hey, I want to use this MCP, do it." What I like to do is I like to check that there is something in the MCP that is perfect to install in Clo code. Okay, they have implemented quick installation for Clo code. And so here, that means if we want to use it in Clo code, we just need to use that line here in Clo code. So let's go back in Clo code here. We paste it. Uh, I can say I will give some context. "Hey, I want to install globally the Playwright MCP." And then I validate. And so it will execute the command, and that's it. It's installed. It's as easy as that. When you want to install validated MCPs, legit ones, it's pretty straightforward. Now, I've seen myself struggling a lot with some underground MCPs, and so I don't recommend to spend too much time on that. Okay.

So now I have my Playwright MCP that has been successfully installed globally, and so it's added to the Clo code configuration. I should find it in this folder if I want to check. Let's keep things simple. And then it can automate web browsers, scrape dynamic web content, test web applications, interact with web pages programmatically. This one is really good when you're developing a UI, by the way. And so it's normally available globally. So I will log that. Yes, let's do the three steps. By the way, in the meantime, I use the command `/mcp`. I can check `mcp status` and I can see Playwright is connected, whatever. And if I use `mcp manage mcps`, now I have to continue in the terminal with just no big deal. And so it will open Clo code in my terminal, whatever. And now I am in MCP, and so you can see here Playwright is connected. If I click on it, I can configure it. Okay. So that's my Playwright MCP server. It's connected, and there are 21 tools. Okay, you can see in Playwright MCP, there are 21 tools. So let's see what is in there. You can see there are many tools: upload files, fill form, install browser, specify for in the config, press key, navigate. Basically, it's all these tools are actions that my agent, my Clo code, can do on a web interface to navigate it through Playwright. Okay. So nothing big here. Just I wanted to show you how to install an MCP really, really quick, and how to check that everything is running fine, and how to check what kind of tools your Clo code has access to through this MCP. Okay. So no big deal here. Let's remove that. Oh, let's clean that properly.

We've covered a lot. Now let me show you how all these pieces fit together into a complete system. Let's make a quick recap. So we've seen that with Clo code, you have several features and also several capabilities that makes it a complete AI-powered system, especially in a business context. Once again, Clo code has been highly pushed for developers and coding use cases, but actually, we are using it for business, and it's amazing. In Clo code, we have basically our folder structure, which represents our different projects or tasks or teams, according to the way you want to call that. I call that either project or my teams. And we have `.clodevrc` files that give for each project, each team, each folder, instructions, mechanisms, ways of doing stuff inside the project. Which means each time Clo code is called, it checks the `.clodevrc` files, and it knows everything about the current work inside this project. So you can consider that as the business brain and the way your business or your business operating system is structured. Then you have your sub-agent teams. They can be available globally or locally at the project level, and I consider that as my different departments, my teams working for me. Okay. And they are self-orchestrated by Clo code, and there are many manners to do this, but let's not enter into the detail. Then you have tools that you don't necessarily see, internal tools. For instance, Clo code has tools for to read files, to do web.

search and so on. Let's not take too much time on this. Let's keep things simple. Cloode manage I mean use the right tools at the right moment without you doing nothing. So it's kind of working in the background.

Then you have the skill which are reusable workflows that are packaged and they can be shared between your teams of sub agents or even between other users like your teammates or your customers or exactly what I'm doing in CCTG. I'm sharing skills some skills with you guys and so all that folder structure cloud.mdi sub agent teams skills tools inside cloud code these are internal capabilities on top of that you have also the possibility to add external capabilities external tools through APIs and MCPs so you could build your own system trying to understand how to use all that and then do trial and errors and so on but there is so much value to having pre-built system so that you don't build from scratch you Don't spend months of work on just building the stuff, designing the folder structure, writing the code.md files, creating the sub aents, testing them, improving them, removing the one that are not working well or building skills library which takes time because each time you're building a skill, you need to test it, you need to improve it. That's that's work, right? Or setting up integration, testing API connection, testing MCP. So you could do this and spend lots of time doing this.

Or you could deploy mine the entire business OS with folder structure cloud.mmd5s mechanism already built in as well as my content team with the 14 sub aents that I have in magnetic content OS the strategic console the AI gross engine plug-in which is yours that that would be the AI growth engine with its framework it structure but apply to your business with the productivity system with the daily planning productivity assessment already built in taking your information what you work on with code on a daily basis to give you assessment about your productivity to help you also prioritize using the brutal prioritization framework on the right stuff for your business. Having AI support to be productive and focused of course would have also my HR team, the GPT creation team, the GPT express mastery OS and skills and scripts library that I develop for myself only focusing on stuff that are useful for business. Of course, you can have access to all this here in the cloud code OS. This is what I'm sharing cloud code OS. I'm sharing my systems, my teams that you can deploy that right away inside your own business and leverage this full system. And of course, you're not alone with just the system. I'm here through implementation calls to help you customize it or make it evolve as you see fit based on your need inside your business. And so that's the power of this code OS program. And so this program is definitely the fastest path for you if you want to have a fully integrated AI powered business operating systems that helps you work on what matters inside your business with the right tool leverage in a maximum manner the latest AI approach. I'm using these systems daily for my own business which mean I update them I improve them and of course all these improvement and updates are transferred to you in the cloud code OS. It's not a static product. It's a living evolving system that is driven by my conception of the business which is all about focus prioritizing the right stuff inside the business solving the image bottleneck and improving the aspect of the business that I call attract attract the right customers convert convert the right customers retain them ensure that they are satisfied with the service delivering outstanding service and ascend them like selling them more stuff I build everything around this like the right focus and maximizing the attract convert retain ascend function or dimension of the business. So we'll talk more at the end about this. Uh but first let me show you how collo code itself operates because understanding this gives you more control.

And here in the next part I will show you some ways to use cl code in a better manner. Uh cool stuff to do with code is use the plan mode. With the plan mode you basically ask cl code to think about what you ask it to do and it will write a plan. And also you can use natural language command like uh sync hard sync harder or hress sync to really push cloud code to use more token and think more about whatever problems you have asked it to solve or whatever task you want it to do. And so this plan mode is a readonly architect mode to to activate it. I will show you you can use shift tab or you can click on the right part of the interface and then cl code will analyze and propose the plan. You read the plan you approve before execution. for huge pl huge task or huge work I have to do with cloud code I always use the plan mode so that I can review its approach I can correct it and sometimes I can better understand also how to implement a given task or a given project and then I approve and then it execute it write the to-dos you've seen that there is a tool that cloud code often which is to-do list that help it be align with what it needs to do and not forget things and so really the plan mode is super powerful it's and I really recommend that you use it each time you're working on significant project so that you have architect code mapping the the plan suggesting the plan and then you can validate and it gives you also the blueprint of what would be done and so that you have no surprise when cloud code implements the advantage of it of course it's safety and control it's also helping you to better think about a given project or given thing you want to implement with cloud code and that's super useful right it's ensure that you have a structured approach you know back in the day when we were working with just prompt or even with just custom GPTs we used to say and that's a prompting technique proceed step by step. Explain to me what you will do first so that I can validate and so on. Here it's same idea but automatically done by cloud code in an optimal manner.

So let me show you what plan modes look like. Okay to activate plan mode you can use here this clicking here you can shift to plan mode. You can use shift tab as well and it would change it. And then I don't have any example here. I'd like to know how I can best explain to CCG member the power of code OS program and what is the value they would get from from it and what they would avoid what what's the pain they will avoid by implementing it and what's the huge benefits they will get if they implement it can you help me with that okay in here I want to use plan mode so that code can think about it find some resources inside my CCG business OS and give me a plan to explain that to to you obviously. So you can see now it's collecting information. Yes, don't ask again. And here the cool stuff is that cloud code won't implement anything. It will just collect information, process that and then come up with a plan based on my request. So here it's gathering context and so on. You can see it has find the information about code business OS value proposition and now it's compiling findings into a clear organized response for me. The plan. So now it's the plan following my query here. So cloud code business OS complete value proposition framework the transformation from chat GP chat session to cloud code business OS with integrated system that knows your entire business persistent memory strategic partner automated operations and intelligence pain point they will avoid morning where was I friction strategic drift weeks lost on low value work what should I work on paralysis okay decision fatigue daily inconsistent productivity with no pattern visibility you cannot analyze your work manual tracking hover okay here we automate tracking blocked without knowing what's blocking you. Hidden dependencies, huge benefits they will get. Time freedom, strategic certainty, compounding intelligence, execution clarity, productivity optimization, project continuity, strategic leverage plus seven integrated workflows. You have the daily operating system, you have the strategic layer, the AI growth engine, you have the infrastructure layer with project cycle, life cycle, workflow, system maintenance, different proof points, implementation path. You can start with the basic feature then work with the strategic feature and then use the complete system positioning for CCG members and key messaging by avatar simple pitch structure fudge ready for you for you use okay so that's a plan and then I can say yes would you like me to create a specific deliverable using this framework including YouTube script sales page copy workshop etc so it's going beyond what I ask but here I have a plan it has gathered information processed it and then suggest me a plan with the different elements I could use and then told me about the resources possibly implementation. So that's that's a a simple example of to use plan mode. Now I can shift back to normal mode or I can just say execute and then it will execute the plan. Plan mode is super powerful to plan your work and be sure that cloud constructure your approach and execute it in a sound manner.

Now let's talk about configuration control. Uh I will cover here the next part will cover settings versus instructions. Let's roll here. What I want to cover quickly is settings.json. JSON and code MD file. We have already covered code.mmd file but there is another file which is settings.json and this is the way it looks like in my CCG business OS. I have a settings.json here and actually in the settings.json you can set up environment variables. Of course you can use code to do it for you but just I want to show you what is inside setup setup setup variables some specific rules that you want to use at the project level here that can be at the full system level. If you go in your user/clode subfolder because in there see the very same settings JSON. Basically the settings.json is in any project you have inside code. Let me show you that. Okay. In my root folder, my system root folder I have also setting.json. There is nothing in there. Why? Because I I don't work globally. I work only in my CCG business operation. In this CCG business operation folder. So here in my settings.json my global settings.json, there is nothing. Okay. In addition to variables and specific rules, you have also permissions and deny. You can see for instance, I don't want anything from this project to write in my AI engine knowledge base. Okay, why? Because I want to keep the knowledge base untouched when I'm working actively on operations. Thing you need to understand is they add some layer of automation in your cloud code system. You have a hierarchy of settings.json. You have the settings.local.json that will work on your local project only for you. That's the one that will be taken into account first. Then if you don't have that one, you have the settings.json that works at the project level, but that can be used by other users as well if they have access to your project. And then you have the the general one that I showed you before, which is this one here in the code subfolder of the root user folder. And this is a settings of JSON that works globally, but it's the weakest. That's the one that will be used the last only if there is no nothing else in the previous one. Just the thing to understand is that the settings.json JSON set machine readable controls in a JSON format that gives tools permission model selection hooks environment variables MCPS and so on. Most of the time you don't need to take care about this even if you can use it to enforce permission authorization and so on but really in a business context most likely you won't pay too much attention to that but I want to mention it anyways and you have the code files which is human readable so that's good for you to understand also what is done in a given project and that of course is exploited by cloud code also to have context when it's starting a new session inside a project or inside a team or inside a department or inside a folder for me folder on your machine can correspond to specific project to specific team and and workflows or specific department like I call them department like in a company you have your department with your teams your sub agents and all that sit inside a folder that can have a clone file that gives context and instructions and a settings.json JSON that that gives some configuration elements and per permissions. They work together. You don't need to pay too much attention attention to the settings.json. I I've spent I mean I've worked hundreds of hours of my cloud code and I'm barely pay attention to that one. The cloud MD file is the most important in my opinion from my experience.

Now, can you have tools everywhere and tools only for specific projects? Yes, there that's the notion of hierarchy. I already covered that a bit when explaining the settings.json. Maybe I did a poor job explaining this to you. So I hope you're not confused but let's try to clarify this notion. Okay, let's talk about hierarchies settings hierarchy from the highest to the lowest. So that's what I tried to explain like few minutes ago. You have local project settings in the code settings.local.json. That's your personal tweet. It won't be available for other teammates. It won't be available on GitHub if you have synchronized what you do in cloud code with GitHub. Then you have shared project settings in the code/ settings.json JSON in all all these dot clothes subfolder are available in all folders you're working on with cl code right it's everywhere in every project or department or teams or workflows you're working on inside a folder on your computer there is a code subfolder and they have settings.json J most of the time there is nothing in there. Side note here. Okay. So the two first layers local project shared project. You can see in the shared project actually it's checked on G on G. So it can be used by other teammates who have access to the same project. If you share a project and then you have the personal global settings in your user folder subfolder.load and file is settings.json and that's your preferences but personal everywhere. An analogy to understand that is that you can have a personal toolbox and that's what is in the clothes subfolder of your root user folder. So that's the tools you carry everywhere personally. Then you have team workshops inside your project in the cloud subfolder of a project folder. And then you have your personal tweaks which is a settings.local.json. Okay. The setting merge they're not replaced but with another priority and the higher priority overrides a specific value. And from what I've experienced and understood, this one for you has more priorities as this one that has more priorities. This one, this logic of hierarchy is true for other object inside code. For agents, for instance, you can have agents in a specific folder. You can have agents in your code subfolder of your user root folder. And if they are just local, they are applied to the project only. If they are the subfolder of the user root folder, they apply everywhere. Same for skill. I know that like currently November 2025 there are some issues about the capacity to invoke skills that are locally implemented. Well, this being said that's a detail. It has not prevented me from from doing what what I wanted to do. So it's a detail but I mention it. Okay. Key mechanism automatic discovery. So when cloud codes work it scans folders and so you don't have to say anything to cloud code. It will just scan folder and see what is available in the settings.json files. And the same with cloud.md files by the way. Some element like skills are autoactivated. And once again skills can be in your user folder. It can be in a project folder right normally they automatically invoked by cloud code. Sometimes I recommend you to enforce the invocation by cloud code by mentioning that you wanted to use the skill. That's the safe path. I would say there is another thing in that I've I've shown you previously without mentioning too much about this. It's command. When you use the slash key if I go there you have access to command. You can see here you can navigate to different commands and you have the slash command here. Okay, like compat content formatter and so on. To be honest, I rarely use the command. But this is your way also to trigger specific command manually without relying on cloud code to invoke automatically skills or other stuff like this. Okay. So the difference I see with skills and command is that skills can be auto invoke whereas common it's you ask cloud code to do specifically something using a command. And I I will tell you more about command in a few minutes. And so this hierarchy of settings.json but also on skills on agents helps you to contain this object or this feature at different level either just for you personally or globally on your system with other people having access to part of your project. For in I've tried to cover configuration control control. This being said most of these I I told you that so that you understand and you you you know that it exists. But really when you're using your system, when you're using cloud code, you don't need to manage that too much, especially if you're using it solo. Now if you're using with other teammates, that's another story.

Now, how can you track what cloud code does? Let's talk about logging. Let's cover logging and tracking. So in cloud code, you have building cloud code logging stuff. Cloud code track tool usage. It logs error and it keeps session history. So if you open a cloud code session, I already showed you that. But if I open a if I check this one for instance here you can see I have my session history and I can open one and I will go on with what I did in the in this given session. Okay so that means you can reuse sessions. So cloud code is keeping history and logging stuff but that's not enough. Why? Because it's mostly technical stuff or just chat session. I've implemented my own custom business logging logic using several mechanism logging different operation in a structured manner and integrating that with different other workflows and processes. This is not part of cloud code by default. This is part of cloud code OS and this is something you can access to in the cloud code OS program. Okay, I'm continuously improving this because what I want to do with the right logging system, I want to be able to track the work that has been done to do some analysis over time about the work that is done about alignment, strategic priorities and many many other stuff for which logging what you do in a structured manner and clever manner is super useful. Really here we are leveraging if you have the right systems and you will have when you join cloud code OS you can really take another step in term of tracking your work planning your work analyzing your work understanding what happened if there is any derail or issues and many many other stuff understanding what teams are doing teams or project or workflow inside load code are doing understanding how to improve that doing some improvement analysis do like the sky is the limit this is what you would like to to have in your normal business with your teammates. Okay, except here you don't need them. It's automated and it's well structured and you can process cross cross project cross team across time analysis and so on. You have a real business journal optimized for business perspective for business objectives and for business use. You can feedback anytime. You can track your progress. You can analyze patterns. Like that's so powerful and that's where real business value sits with such a system as the cloud code OS because now you know exactly what you what cloud code has worked on but also what you are your teammates if you put them in the system have worked on you can spot patterns you can check efficiency of workflows you can find bottlenecks you can fly find productivity derail you can find like many stuff you can do audit tray for compliance you can you use use it for continuous improvement and the cool stuff is that all this is is done automatically and that's super powerful. If I show you an example for instance if I check yesterday I did a productivity assessment. So you can see here I can come back to the the right chart session and that's my productivity assessment and look what it has spotted pattern identified infrastructure building crowds out execution for two consecutive days. Why? because I had set business priorities I should have worked on last Saturday and actually I spent some times creating what he calls infrastructure which are basically systems instead of working on the business priorities I worked on infrastructure and so it has spotted this because obviously I haven't been able to address some priorities I I needed to work on and it's analyzed why I did that and did some recommendation which is hey it's okay to work on infrastructure but do it after the one thing after working on the main topics you you have planned for the day and it was able to do this because it track everything I do and it understand the strategic environment strategic how to say that priorities and also the different projects I'm working on the different task I'm working on okay can see the power of it that's crazy anyways sometimes you want to trigger specific workflows manually and so that's when you use custom commands it's so we have already covered really quickly custom commands in the previous chapter but let's take a closer look to this you already have pre-built commands in cloud code and I already show this if you type slash you have pre-built commands or all that okay but you can also set up your own commands which are basically workflows and processes simple processes that you can trigger manually so that code could code could execute that and so what are custom commands there are user initiated multi-step workflows they are stored in once again in a doc code folder in a subfolder that is commands and you can invoke them just using the command name some analogy it's a bit like a macro in excel or a shortcut on your phone it's just a a sequence of preparation that you can trigger anytime. And the difference between skills and command because that might be a question you have is that skills are auto invoked by code. They're automatic context based which mean cloud code understand the context and identify the skills that match the context and then we use the skills and use progressive disclosure go deeper if necessary take some resources that are inside the skills maybe a script that is inside the skills and execute. Whereas custom command it's you invoke manually and trigger explicitly something that cloud code has to do. And so typically it's very useful for recurring workflows. For instance, if you want to do weekly planning, you can have a custom command to do this. It's also useful for complex sequences when you have multi-step processes to do. And when you want to standardize process and printer them manually with full control and so how you can create that commands, custom command, you use cloud code and you ask for custom command. Okay, you explain what you want and cloud code will help you create the custom command. And basically what cloud code will do, it will create a file that it will put in the code/comands subfolder. Once it's done, you just call the custom command with the slash and the name of the command. Okay. And if I show you what a custom command looks like. Let me find one for you. In my CCG business operation in my code folder, I have here a command which is in it business. And you can see like skills like sub aents. It's a markdown filed file with a YAML header metadata with a description. And this is what this command does. It init initialize business project cloud.md with business specific guidance. It has some tools it can use read write edit globe which are the basic tools of cl code and then just the md file instruct what cloud code should do when I trigger this command init init business. It's a custom command I have created so that I have cloud code creating a project infrastructure with a cloud. MD file that is specifically structure for what I want to do. So I want cloud code to do something in a specific manner and I want to trigger it manually. I use a custom command.

Enough with commands. Now how about syncing codes work with external system. Interesting, right? So let's talk about GitHub and how to do backup version control and collaboration. So GitHub is really for coder. But guess what? We are using cloud code not for code. We are not developers. We are just normal people, entrepreneurs who have businesses and want to leverage the power of AI to run our business in a lean, effective and focused manner so that we can get free right okay so we I'm using GitHub to for different purpose to back up the work I do inside my business OS of course to control the version to be able to revert back in a more automated manner if need be and also to sometimes collaborate with other people right so what is GitHub quickly it's kind of a dropbox for code with time travel. So you can back up stuff, you can control version, you can collaborate, you can time travel that means you can revert back what you did and there are lots of many other stuff but for all purpose that's the main stuff to understand the main the main features of GitHub to understand there are many other stuff you can do for GitHub. It's very powerful tool for coders but for us it's good enough. You can work in parallel and test different approach in parallel to check the ones you prefer and then merge the one you prefer on the main project like you can do so many stuff whatever the advantages for business is that you can recover for any issues that's a backup okay which means you can also use your system on another computer if need be right why like let's suppose that your your computer is dying because one of the problems code is that you work on local file right and what if you have a virus or what is your your computer is dying now you're screwed not any longer if you have a backup system on GitHub Because now you can open another computer, log inside your GitHub account, import everything inside your new installation and you're good to go. You can also control the version history so you can see the system evolution. You can collaborate with other people, experiment safely. You can audit stuff, understand who work on what and who change what. If you have several people working on the same cloud code business OS and it gives you remote access because you can access from any machine as well. And so how it work, you ask cloud code to create a GitHub repo. You connect close code to GitHub. It's something you do once and then if you give access to cloud code to the GitHub CLI which is some sort of module, cloud code is able to create whatever repo in your GitHub account. And so that makes stuff way easier. And then of course cloud code can commit changes which mean push changes on your GitHub repo and and commit them so that now they're alive. You can also syn synchronize between your local cloud code business OS and what is inside GitHub. You can clone, revert, collaborate anytime. That's super powerful. That's really easy to set up. You ask code and it at some point it will ask you to connect to GitHub with your your authentification method. You you do this. You accept connection. You accept the the synchronization the connection between cloud code and GitHub and you're good to go. That's really straightforward.

Let me bring it all together because we have covered many things here. And let me show you what's next. So we have covered a lot. We have covered filebased memory and system prompt with cloudnd files. We have covered sub agent organized as teams that allows you sequential or parallel execution having the sub agent auto orchestrated by cloud code. We have covered skills with our usable reusable workflows which are standard operating procedure that you can equip your cloud code system with or that you can also equip your teammates using cloud code with so that you can enforce SOPs at the company level. We have covered parallel execution whereas it is with sub aents working in parallel or tools that cloud code calls in parallel. We have covered access to external tools through API integration and MCPS. We have covered some feature that makes the work of cloud code more effective like plan modes some commands. We have covered the difference between settings with settings.json files that are more for machine reading and cloud.md instruction. well fit for you to understand what happens at the project level and also for code to load context when it start a new session in any project. We have talked about hierarchy how you can organize skills agents and settings.json files at different levels so that it can be used by only you locally on your machine everybody working on the given project or globally on your system. We have covered logging mechanism quickly that is super useful for many business use cases like productivity assessment, planning, strategic alignment checking and many other stuff. We have covered custom commands quickly how you can set up small automation or specific workflows that you can trigger manually which is a bit different from skills where skills skills are triggered automatically invoke automatically by cloud code. Custom commands helps you standardize some stuff you do and trigger them manually. And then we have covered backup and version control using GitHub in a very quick manner. So you can see that when you understand all these features, you have a complete system. But actually that's just step one. Knowing the features, understanding how they work and where you can access them. It's only step one. Step two, it's implementing a cohesive system that combine all these feature in a smart manner for a business context in order to increase business efficacy. in order to increase focus on what matters inside the business and in order to increase the power of execution your business leveraging AI powered system the best manner of course you can make your own trial and error and spend weeks maybe months to try to set up the system or you can use a pre-made one and that's exactly what I give you in cloud code OS inside CCG community in the cloud code OS you have the entire business OS that's the one I use for myself so you can be sure that it's topnotch and I make it better consistently when you have my business OS you have the folder structure that is directly applicable to your business. You have the cloud MD files which has the system prompts and a smart logic for the different folder and different project. You have an already integrating mechanism for productivity for planning for cross project analysis and so on. You have also my own teams my department. You have my content team with the magnetic content OS that has more than 14 sub aents to work on any type of content on LinkedIn, Twitter, YouTube, Instagram, etc. Working on hooks, working on Sundays, working on all that stuff related to content, applying the content framework that I teach also in CCG. You have what I call my strategic console which is the AI growth engine plug-in with all the frameworks to really describe in a strategic manner your offer your target avatar your strategy your funnels your different products and so on. You have my productivity system with the daily planning with the productivity assessment with the logging systems to log the different projects to log the different tasks to log the different change that you do with cloud code inside your business environment. You have of of course my HR department, what I call my HR department, which is the GPT creation team, the GPT Express Mastery OS that use my GPT Express mastery framework to craft custom GPTs, but also some agents but at a superior level. way better than what you would do with just cloud code or just code or just ch why because it's using a method that I've experimented with for more than two years and that I talked to many people who just can tell you that it's a crazy good frameworks to craft superior custom GPTs and AI agents but there is more you will also access to essential skills like the skills factory for instance the skills to build other crazy good skills and other skills that I develop based on business use case and business needs you also have some scripts library to connect with external tools and you just need to put your own API keys and you're good to go and there would be much more because once again this is the AI part business system to grow lean GPT business a business that is optimized with lean processes to focus really on what matters to grow the business implementing the best approaches to attract convert retain ascend customers without having to spend months to learn them that would be much more. I will update this based on how I use the system and how other member in the cloud code OS are using the system to get results and also based on the new stuff that entropic will release for cloud code because guess what clothes is a big bet that entropic made. This is a tool they are investing investing massively to make better and there are so much great stuff that will come in the future and that will be updated inside cloud code. Here the key difference is that you're not building anything from scratch. You can just take the the resources I share with you. These are folders basically that you put on your computer and you're good to go. You open cloud code in your Visual Studio Code interface and you go in the right folder and good to go. Everything is there already set up, right? So you can deploy in minutes what I've built, tested and refined for days, weeks, months. If you join the code code OS program, you have my support through implementation group calls that we do and also you have access to me through DM and so we can discuss about your specific case and I can help you tweak a bit the pre-made system that already there and that you can deploy in minutes. It's a living system. I update it constantly and my updates are directly implemented in cloud code OS so that you can benefit from them and of course you may have some specific requests and if they all make sense I can implement them as well which mean here we are leveraging in cloud code OS we are leveraging collective intelligence to make our business OS better with time so finally if you're business owners whereas you have online business where are you you're a coach you are service provider you you do consulting or you need to create content or you do product development this is for you. If you want an AI team to end hand end hand end hand end hand end hand end hand end hand end hand end hand endle execution inside your business so I can focus on strategy and you can just orchestrate your teams your department as a CEOs that's for you because this code OS is the business operating system that you can use right away to have these teams working for you and you can treat it also with the support inside the community and so if you want proven systems that are ready to deploy immediately you can learn more about this in CCG my school community and here in cloud code OS this is the module where everything happens and where the resources to have your own AI powered business OS inside cloud code are sitting waiting for you. So, I hope you enjoy this Toad Code Mastery for Business course and I see you inside CCG.