Google is taking a major leap into the world of AI agents with the introduction of Agent Mode for its Gemini platform, announced during the Google I/O 2025 developer conference on May 20.
Agent Mode, powered by Google's Project Mariner technology, transforms Gemini into an autonomous assistant capable of performing complex multi-step tasks with minimal user oversight. The feature enables Gemini to browse websites, interpret visual and textual elements, and execute actions on behalf of users.
"Imagine simply stating your objective, and Gemini intelligently orchestrates the steps to achieve it," Google explained in its announcement. "Agent Mode seamlessly combines advanced features like live web browsing, in-depth research and smart integrations with your Google apps."
In a demonstration during the keynote, Google CEO Sundar Pichai showcased how Agent Mode could streamline apartment hunting. The AI can search listings on Zillow based on user criteria, adjust filters, and even schedule property tours—all without requiring the user to visit multiple websites or manage the details manually.
The technology operates on cloud-based virtual machines and can manage up to ten tasks simultaneously, allowing users to focus on other activities while Gemini handles time-consuming processes in the background. This capability represents a significant advancement in how users might interact with the internet in the future.
In a move toward industry standardization, Google also announced compatibility between its Gemini API/SDK and the Model Context Protocol (MCP) introduced by Anthropic. MCP is an open standard that enables AI assistants to connect with external tools, data sources, and systems through a unified protocol.
"MCP is a good protocol and it's rapidly becoming an open standard for the AI agentic era," said Google DeepMind CEO Demis Hassabis, highlighting the company's commitment to developing it further alongside Anthropic and other industry partners.
Agent Mode will initially be available to Google AI Ultra subscribers in the United States, with plans to expand to more countries soon. The company is also bringing Project Mariner's capabilities to developers via the Gemini API this summer, with trusted testers like Automation Anywhere and UiPath already building with the technology.