Google announced during Google I/O 2025 that it is rolling out the project mericious, the company’s experimental AI agent that browses and uses websites to more users and developers. Google also says that it has been quite updated how the project mergeer works, allowing the agent to take about a dozen tasks at a time.
Google’s new $ 249.99-The US subscriber project for the AI Ultra Plan will get access to the merge, and the company says that support for more countries is coming soon. Google also states that this project is bringing merrinner’s capabilities to Gemini API and Vertex AI, allowing developers to manufacture applications operated by agents.
The first unveiled at the end of 2024, the project marinner represents Google’s boldest effort, so that users interact with the Internet through AI agents. At the launch, Google Search leaders stated that they see the project merrinner as part of an original user experience shift, in which people will hand over more tasks to AI agents, rather than going to websites and completing those tasks.
For example, project mericious users can buy tickets for a baseball game or ever buy grocery items without going to a third-party website-they only chat with Google’s AI agent, and it seizes websites and takes action for them.
Project Meriner competes with other web-growing AI agents, such as Openai’s operator, Amazon’s Nova Act and computer use of anthropic. These devices are all in an experimental phase, and the experience of techcrunch has proved the prototype to slow and prone to mistakes.
However, Google says that the project has been taken from initial examiners to improve the capabilities of the merrinner. A Google spokesperson told Techcrunch that the company updated the project merrinner to run on virtual machines in the cloud, such as Openai and Amazon agents. This means that users can work on other projects while Project Meriner completes the task in the background – Google says that the new project can handle 10 tasks simultaneously.
This update makes the project merrinner much more useful than its predecessor, which went to the user’s browser. As I mentioned in my initial review, the initial design of the project merrinner meant that the users could not use other tabs or apps on their desktop while the AI agent was working. Defeated the purpose of such an AI agent – it will work for you, but you could not do anything else when you were working.
In the coming months, Google says that users will be able to use the company’s AI-operated Google search experience AI mode. When it is launched, this feature will limit the search labs, Google’s opt-in testing ground to search features. Google says that it is working with the ticketmaster, Stubhub, Rage and Vagaro so that some of these agents can get electricity.
Different today, Google unveiled the initial demo of another agent experience called “Agent mode”. The company says that this facility combines web browsing with other Google apps along with research facilities and integration. Google says that ultra subscribers will soon get access to agent mode on the desktop.
In this year’s I/O, Google is finally ready to ship the experiences of the agent that it has been talking about for years. Project mergers, agent mode, and AI mode are all ready to change how users navigate the web, and the sellers talk online with their customers. Web-buzzing agents have great implications to the Internet economy, but Google is ready to exclude all these agents in the world.
