Chrome Codex 5.5 Unlocked: Enable It Before Anyone Else in Europe!

In this preview video, I'm going to show you how to get the latest Codex Chrome extension. ChatGPT Codex 5.5 can now use Chrome directly on macOS and Windows, and that changes everything. In this tutorial, I'll show you how real estate agents can automate their work by directly asking ChatGPT Codex, with the Chrome extension, to search for properties and contact professionals or sellers, delegating real work to ChatGPT with the Chrome extension. I'll also show you how to use the ChatGPT Codex Chrome extension directly in NotebookLM and whether it's possible to integrate NotebookLM's core functionality directly into ChatGPT. That's exactly what I'm going to show you in this video. I no longer need to store RAGs locally. I no longer need to store them on Obsidian. Thanks to the combination of NotebookLM with Gemini 3.1, I can directly retrieve all the data from the interface; it retrieves the information and feeds it back into the response. How do you install it? How do you get early access in Europe? Yes, because it's currently only available in the United States, but I'm going to teach you how to configure it, get early access in Europe, and show you its potential for businesses that can automate part of their work today using ChatGPT's new Chrome extension. A whole new world is opening up: the world of automation for every business. This tutorial will show you how. To install Codex early, go to this address, which I'll include in the lessons and, of course, within the AI ​​basics section. By clicking "Access," you'll get the direct link to connect to the Codex Chrome interface. And of course, you already have it within the training materials, with the prompts we'll be using in this section. How do you display this extension? First, activate a VPN with a US connection. You can use any type of VPN. Once you're logged in, you'll be able to display the extensions section. Once it's displayed, pin it to the menu and click on settings. This will allow you to synchronize Google Chrome with Codex. Here, you have all the session management approval rules with Codex. The first option lets you choose whether Codex should request permission before opening a website. So, you choose whether it has the right to do this automatically or if you want it to request it. Does it have the right to access browsing history? Can it download files from the website or not? It's up to you to configure these settings. And be aware, I'll show you in the next part of the video how to also modify the Codex system prompt. The system can also send information about websites. In this specific case, I advise you to leave it under manual control. You have the option to block domains, allow others, and specify which download domains are permitted and which are blocked. OpenAI has thought of everything with the Chrome extension, including a number of security settings, and we can now truly imagine Codex taking complete control of the workflow. Once this modification is made, you need to keep the "connected" option enabled in the interface, and I'll show you one of the possibilities we now have in the structure. When working in the interface, to call Chrome's function, you'll type `@Chrome`, and you'll now have the "Call Chrome" option. From there, you'll issue work instructions. So imagine that when you give an instruction, there are steps. Step one: go to the Google page. Step two: if you give it work, you'll need to define prompts and functions. In this context, Chrome is much more... It's more than just a browser. It's an intelligent interaction system that allows you to work on the interface. So, I have what's called a system prompt. When I start my session, my workspace is prepared. Let me explain why. Most accounts today use login credentials. These credentials are your passwords, and they need to be secured. To secure them, I'm going to give you the prompts to use to protect all your passwords when you use Chrome within Codex. Here's why. To view the window, we'll ask Codex to display it within the Codex interface. We now have access to the interface and therefore to the workspace. Every time you log into accounts, you'll have passwords, and Codex can store them for you, but you'll need to secure them. Here's the advice I'm going to give you. Create what are called environment variables and store them in the format that defines URL, account, email, and password. You define security rules, usage procedures, and allowed and forbidden exits. The environment variables you use to connect to interfaces must be stored and protected on your computer. If you want to learn how to use AI professionally and scale your work, go to "Mastering the Best of AI 2026." You'll learn to become a certified entrepreneur who automates their business in less than 15 days. And AI today is an extremely dynamic field. You have all the updates included. We are number one in terms of updates. You have more than 85 hours of lessons that you can complete at your own pace. All the main AI models are included in this training. You have ChatGPT, ChatGPT Codex, AI agents, and of course, Claude, Claude Cowork, and Claude CLI. You're going to become a pro with a clear path. A truly professional level, you'll learn to feed an agent with your data in a usable, clean, and useful way. Just go to the training courses in the description. So, the prompt you have here, I'm going to give it to you, and you're going to integrate it into the advanced settings. You click on settings, you go to the customization section, and that's where you added the instructions for password management. You go to your section and add the formatted instructions with the access and environment paths and access conditions to the system prompt. You save, and from now on, you'll be able to teach Chrome, using Codex, to use your work interfaces. Here's what this changes: If I now need to work in the interface, if I now need to work in real time within an interface, well, I can simultaneously be in the work interface and ask Codex to take control of the screen. So, I'm going to tell you something that's incredibly interesting in this specific case. I've shown you that it 's possible to create what's called a second brain. Imagine that from now on, you no longer need to use your entire data plan when your AI searches your database. I'm going to show you a very interesting trick: using the NotebookLM interface. NotebookLM, as you know, is a Google framework that allows you to keep all your data in a closed environment. So, it has the advantage of working exclusively with your own data. Now, when you create a database, most of the time you're forced to use a very large portion of your AI's context capabilities because you have to send a huge amount of data to the interface. But imagine for a moment The ability to combine a RAG retrieval search with NotebookLM, called remotely via ChatGPT Chrome, is a significant advantage. Let me explain: when you create a database, the costly aspect is having an AI that reads your documentation, which takes time to retrieve the data. But what can be combined with this type of system is far more powerful: the ability to use NotebookLM as a search engine, capable of simultaneously reading images and text and serving as a knowledge base. Imagine being able to transfer your entire knowledge base from Obsidian, send it to NotebookLM, and from now on, thanks to the ChatGPT 5.5 Chrome extension, let Codex take control of the interface, perform RAG searches, retrieve the information, and return it to your response interface. This is precisely what the model will do completely automatically. I no longer need to store RAGs locally. I no longer need to store them on Obsidian. Thanks to the combination of NotebookLM with Gemini 3.1, I can directly retrieve all the data from the interface. It retrieves the information, feeds it back into the context response, and you've created a very lightweight RAG system extremely quickly, where all the data modeling is handled directly by NotebookLM and therefore Google. The model retrieves the response generated by NotebookLM, and you now have a combination, thanks to Chrome, of a chatbot and the DeepSeek interface. Now, I'm going to show you that we can take this system even further: we can completely automate the work. We can now close this interface, create a Skill function, and you'll see what happens. We'll use a Skill function that will launch the ChatGPT Chrome function, ask the questions in the interface, and retrieve all the data without opening any other elements in the Codex chat. So, I'm leaving this visible so you can see that we can use this part of Chrome to create a standalone browsing application, and you can retrieve information in real time. The system works, retrieves, and feeds it back into the conversation, and you continue your work. Imagine being able to create RAGs (Requests for Assistance), respond to clients, customer service databases, or mailing lists by using different Chrome windows that retrieve the data, process it, and respond on your behalf because you'll automate it through Skills. That's exactly what we've just done. We've converted the Chrome function into software capable of retrieving data, and ChatGPT, thanks to this Chrome extension, can work and interact with the real world within it. There you have it. If you'd like to see how I created this Skill application with the Chrome interface, feel free to let me know. I might include it in the training. In the training, you'll also find the protection prompt for entering all your environment variables. And I have to say that this Chrome tool in ChatGPT is really opening up some exceptional possibilities. We've been talking about knowledge base storage. So here I am on a standard page where I would normally type my questions. I'm taking control of the interface. But imagine other professions, like real estate agents, where previously an agent had to go to each listing individually to gather information, get phone numbers, energy performance certificates (EPCs), house prices, and contact people. Imagine now that we're going to create a Skill. We'll just give two instructions to see what's possible in ChatGPT, but it opens up absolutely incredible possibilities in the world of automation. It's about starting to truly delegate... The task: go to ad number 1, ad number 2, extract the price, the energy performance certificate (EPC), the number of bedrooms, the surface area, the presence of a garden (using a Boolean function), and summarize the information in a table. So, we're giving a relatively simple instruction. We're only considering two cases and we're going to ask it to open the app visually. So, it regains control of the screen and goes to Le Bon Coin. However, some devices might be detected. So, in this specific case, it worked. I haven't given any information about the captchas. I'll see how it handles them, but it does. Look, continue without accepting. So, that's pretty good. There might be a slight delay. Can I retrieve the first two ad links? So, let's say click, we'll give it the information. This is what we call HITL (High-Intensity Learning to Read), meaning we have to code instructions for the model so it knows how to behave. So, in this specific model, what's expected is for it to open the first listing, retrieve the information from the interface, close the page, and return to the previous page. If it doesn't do this, that information needs to be entered into a Skill, within the expected behavior of the model. Therefore, we need to design AI today not as a chatbot system, but as a work system where we delegate tasks. So, in this test, we did a completely live test. Nothing is pre-prepared. I'm showing you this because I saw I could get access earlier. So, I'm showing you how it will work. But imagine tomorrow a real estate agent who spends hours gathering information on properties or sending messages to clients. I've already done this with automations using Claude and Chrome. We just retrieved information about the houses, energy performance certificates (EPCs), number of bedrooms, surface area, and garden. Now we don't need to spend so much time on certain tasks and we can automate them quite easily. If you'd like to learn how, leave a comment and I'll create a more advanced tutorial. I'll leave you the prompt in the training materials to protect your system from access codes because starting tomorrow, you'll be able to begin automating your work using ChatGPT Chrome in Codex.