Google’s Project Mariner

What is Project Mariner?

Project Mariner is a research prototype that explores the future of human-agent interaction, particularly within web browsers. It automates tasks such as online shopping, information retrieval, and form-filling, aiming to enhance user productivity by delegating routine web-based tasks to an AI agent.

https://deepmind.google/models/project-mariner/


⚙️ Key Features

  • Multimodal Reasoning: Mariner observes and interprets various web elements, including text, images, code, and forms, to understand the context and plan actions accordingly.
  • Teach and Repeat: Users can demonstrate a task once, and Mariner will learn to replicate the workflow for similar future tasks, enhancing efficiency over time.
  • Simultaneous Task Handling: The agent can manage up to 10 tasks concurrently, operating in separate browser instances running on virtual machines.
  • User Control: Mariner keeps users informed of its actions, allowing them to prompt or halt the agent at any time, ensuring transparency and control.

🧪 Real-World Applications

  1. Online Shopping: By accessing a family recipe from Google Drive, Mariner identifies missing ingredients and adds them to a shopping cart on Instacart.com.
  2. Task Management: The agent can navigate to an email inbox, find a travel plan, and book a ticket

🧑‍💻 Developer Integration

Google plans to integrate Project Mariner’s capabilities into the Gemini API and Vertex AI, enabling developers to build applications powered by the agent. This integration will allow for the automation of complex workflows, such as form-filling and data collection, within custom applications.


📌 Access and Availability

As of May 2025, Project Mariner is available to Google AI Ultra subscribers in the U.S. The company aims to expand access to more countries and integrate Mariner into Google Search’s AI Mode, currently in the Search Labs testing phase.

📺 See Project Mariner in Action

For a visual demonstration of Project Mariner’s capabilities, watch the following video:

Leave a comment