Today we want to give you a preview of Browser Operator, an AI agent that can get stuff done for you in the Opera browser. This feature we’re currently testing marks the first time a browser can use AI to perform tasks on the web and allow you to do more meaningful things with your time instead. Browser Operator is native to the browser and makes it easy for you to remain logged in to your websites. This also makes it safe because login information isn’t sent to any third party, and additionally makes the process faster than a server-based solution. What’s more, Browser Operator does this in an efficient and intuitive experience, where you are always in control. 

We currently find ourselves at a crossroads where browsing and AI can finally meet in a meaningful way and this Feature Preview shows what that may look like.

Browser Operator takes AI out of the chatbot and integrates it directly into the browsing flow. This aligns with the broader role of the browser: it’s there to make you more powerful while you’re online, browsing the web, and to give you the tools to use your time more effectively.

Let’s take a look at the example from our Feature Preview video: buying socks is something most of us have to do at some point, but not something that many are particularly passionate about. What if you could outsource that? Letting Opera’s Browser Operator do that for you could free up your time to meditate or do something else more meaningful to you. These are the types of questions we’ve been asking ourselves while testing the Browser Operator in this and many other contexts. This future we’re describing is near and we’re here to show it to you.

So how does it work? Browser Operator understands your written instructions in natural language using your browser client’s local resources, and our AI Composer Engine processes this information to complete tasks in the browser. Browser Operator will pause and ask you to take action whenever it needs you to fill in a form, confirm an order, or perform some other sensitive task. In this entire process, you are in control and can take over at any moment from Browser Operator, while you can also review all the steps it took to perform the task you requested. 

What sets the Browser Operator apart

This Feature Preview is a proof of concept in which we’re taking advantage of our own expertise in making web browsers and in working with AI in a way that matters to our users. We know how a webpage is built and what it’s meant to look like – so we brought AI into our own playing field and created Browser Operator. 

Our approach to Browser Operator focuses on empowering users while preserving their privacy. As opposed to other solutions that are being tested out there, our concept of an AI agent in the browser doesn’t rely on screenshots or video capture of the browsing session to understand what’s happening in it – nor is it a version of the browser running in the cloud with your user credentials. Opera’s Browser Operator runs natively inside your browser, on your device. It uses the DOM Tree and browser layout data to get context – meaning that it uses a textual representation of the webpage.

This makes our solution faster because the Browser Operator doesn’t need to “see” and understand the screen from its pixels, or navigate it with a mouse pointer. Another advantage of that approach is that the Browser Operator can access the whole page at once (in most cases), without the need of scrolling through, effectively reducing the overhead and time needed to bring a task to completion. 

The best part is that Browser Operator works in the same environment as you: the browser. It doesn’t require a virtual machine or a server in the cloud. This also means that your browsing history, log-ins, cookie settings, etc., are being kept locally in your device, making the user experience smooth and private. Since operations are performed locally, the user can elegantly shift between the operator control and user control. 

Since the Browser Operator sees the webpage data the same as the browser, that means that the popup dialogs – like cookie acceptance and verification dialogs – don’t represent an obstacle to access the content of the page. Browser Operator can do this because it can interact with elements in the webpage that aren’t visible to the user.

AI is changing the way you browse

At Opera we have showcased to the world how AI can be integrated into a browser through several initiatives. We were the first ones to integrate AI-powered features into the browser and we also developed our own built-in browser AI – Aria. For the past year we’ve been improving Aria with our AI Feature Drops program, which gives you more features and AI functionalities.

We’re constantly thinking about how to empower your browsing through AI, and we’re committed to continue innovating in this space.

The Browser Operator user experience

While currently in Feature Preview, Browser Operator is accessible from the Opera sidebar and from Opera’s Command Line. To use the Browser Operator, just write what you want it to do for you.

For example, you can buy and order things online, book tickets and events, or even collect information from a website to fill in a spreadsheet or a document. 

Browser Operator will let you see what’s going on throughout the process as well as what steps it took to complete the task at hand: 

When you give the Browser Operator a task in the form of a prompt, it will work to complete it and occasionally may need your input to finish it – this is what can be considered as having a “human-in-the-loop”. You can then interact with the webpage directly or provide more information through the Browser Operator’s chat.

For example, if you need to fill in a form you can either type the information directly into the webpage or give the information to Browser Operator and then click on “Continue” to let it resume and complete the task. 

Additionally, you can cancel the task that the Browser Operator is performing at any time by simply clicking on the cancel button:

Finally, when the whole task is complete you can either end it there or give further instructions to Browser Operator to elaborate on the task that it just finished. It’s polite to say ‘thank you’ 🙂 but for example, if you placed a wrong order, you can instruct Browser Operator to cancel it. Then, Browser Operator would understand which order you’re talking about and cancel it for you. 

A private and secure approach to AI browsing

When performing tasks for you, Browser Operator needs to understand the context of what you’re doing, and to do so it utilizes two things: the instruction you give it as a prompt and the relevant information from the webpage. These are the only two things that will be processed by our AI and nothing else, ensuring the privacy and safety of your data when using Browser Operator. 

And now you might be wondering what happens when you have to type in your credentials, phone number, or even card details when utilizing Browser Operator. Well, here’s the important part: We developed Browser Operator so that no information – no keystrokes, no screenshots, etc – will be sent to our server when you use it. When Browser Operator requests you to intervene, or when you take over, no information is being processed by the AI, and it will only start reading the website again when you click “Resume”.

For example, if you’re ordering socks in a store, if you have to input your card details and address, whatever you type in the webpage will not be sent to us – only to the webpage you’re interacting with. Additionally, if you’re already logged-in on a website, Browser Operator won’t be able to access your profile details.

Opera’s path toward agentic browsing 

Throughout our history we’ve been pioneering functionalities that then become the new standard in the browsing world: we invented tabs with Opera 4, we also were the first to introduce search in the address bar, as well as the first one to offer a built-in VPN and messengers in the sidebar and the first one to offer native browser AI. In 2023, we redesigned our flagship browser to make it AI centric. Since March 2024, we’ve been pursuing our AI Feature Drops program, through which we test experimental AI features such as local LLMs, Image Generation, and AI Tab Commands in the developer stream of the browser. Today, we’re redefining the role of the browser yet again, by becoming the first browser to perform tasks for you, thus becoming agentic.

Availability

Browser Operator is currently available as a preview of the upcoming functionality. We expect to launch the new Browser Operator as part of our AI feature drop program in the near future.

AI AgenticsBrowser OperatorOpera AI

---