Google just declared at Google I/O 2025 that they are launching Project Mariner, their experimental AI agent that cruises the web and interacts with websites, to more users and developers. They also mentioned that they have made significant updates to how Project Mariner operates, allowing the agent to handle almost a dozen tasks simultaneously.
Who’s getting access to this shiny new technology, you ask? Well, U.S. subscribers to Google’s fancy $249.99-per-month AI Ultra plan will be the lucky ones. Google promises that they will soon expand support to more countries, so everyone can get a piece of the action. Additionally, Google is integrating Project Mariner’s capabilities into the Gemini API and Vertex AI, giving developers the tools to create applications powered by the agent.
Project Mariner made its grand entrance in late 2024, marking Google’s most ambitious attempt yet to revolutionize how people interact with the internet using AI agents. Google Search leaders envision Project Mariner as a key player in a fundamental shift in user experience, where individuals will entrust more tasks to an AI agent instead of navigating websites themselves.
Now, let’s talk competition. Project Mariner is up against other web-browsing AI agents like OpenAI’s Operator, Amazon’s Nova Act, and Anthropic’s Computer Use. These tools are all in the experimental phase, and from what TechCrunch has seen, they can be a bit sluggish and prone to errors. However, Google claims to have taken feedback from early testers to enhance Project Mariner’s capabilities.
Not really sure why this matters, but Google revealed that they revamped Project Mariner to operate on virtual machines in the cloud, similar to agents from OpenAI and Amazon. This means users can multitask while Project Mariner handles business in the background — Google boasts that the new Project Mariner can juggle up to 10 tasks simultaneously.
In a move that significantly boosts Project Mariner’s utility compared to its predecessor, which ran directly in the user’s browser, Google is allowing users to access Project Mariner in AI Mode through Google Search. This feature will initially be available in Search Labs, Google’s testing ground for search features. Google is collaborating with Ticketmaster, StubHub, Resy, and Vagaro to power some of these agentic flows.
On top of all that, Google teased a preview of another agentic experience dubbed “Agent Mode.” This new feature combines web browsing with research features, integrations, and other Google apps. Ultra subscribers will soon have access to Agent Mode on desktop. It seems like Google is finally ready to roll out the agentic experiences they’ve been talking about for ages. Project Mariner, Agent Mode, and AI Mode all appear poised to shake up how users navigate the web and how businesses engage with customers online. Web-browsing agents could have significant implications for the internet economy, and Google seems prepared to unleash these agents into the world.
In conclusion, Google’s Project Mariner is set to revolutionize the way we interact with the internet. With new capabilities, expanded access, and integration with other Google services, the future looks bright for AI-driven web browsing. Get ready to delegate tasks to your friendly AI assistant and enjoy a smoother online experience. Exciting times lie ahead in the world of technology!
