Google quietly updated its list of user-triggered fetchers to include a new one called Google-Agent. The new agent will be used by its Project Mariner tool that began as an AI browser agent and may now be part of a pivot to compete with OpenClaw-style personal agents.
OpenClaw
OpenClaw is a new type of personal AI agent assistant that is able to perform a wide range of tasks online. In fact, it is able to form teams with one agent as the manager (the orchestrator) handing out tasks to specialized agents in a team. These AI agents run from a laptop or desktop device as well as in hosted environments. They are model-agnostic and can connect to any cloud-based AI providers like Anthropic (Claude), Google (Gemini), and OpenAI.
Chinese AI providers like MiniMax, Moonshot AI (Kimi), Alibaba Cloud (Qwen), and DeepSeek are increasingly popular because they are significantly less expensive than mainstream American AI providers, further driving the personal AI agent boom.
The personal AI agent space is so popular and important that OpenAI hired the developer of the OpenClaw AI agent, Peter Steinberger.
Google’s Project Mariner
Project Mariner was announced in 2025 and was only available to Google AI Ultra subscribers and others who were allowed in as Google Labs testers. The initial version of Project Mariner was a browser style assistant that you could tell what to do, and it would go out onto the web and accomplish various tasks.
A video of Project Mariner in action showed it to be a fairly clunky way to navigate the web, with one tester calling it “far from perfect.”
Project Mariner Test Drive Video
Pivot To AI Agents Called LAMs
AI agents are exploding right now, largely in the developer community, especially as it intersects with the vibe-coding trend. AI is currently used for building software, WordPress plugins, creating blog posts, and monitoring and posting to social media. AI agents are essentially robot workers that can do all of that autonomously.
These kinds of user agents are becoming known as Large Action Models (LAMs). A LAM understands what a user wants accomplished, breaks the goal up into steps, clicks buttons, calls APIs, and carries out tasks autonomously or with human oversight. Unlike LLMs that basically say things, LAMs actually do things.
The imminent release of the AI-friendly WordPress 7.0 may usher in a period of rapid evolutionary change in how businesses create and manage websites, and AI agents will quite likely expand and play a big role in that.
Many of the Project Mariner staff have moved over to the Gemini Agent product, as some of the capabilities and insights from Project Mariner are moved over to other projects. Wired reported that it received confirmation, one day before Google announced the new Google-Agent crawler, that Google was moving Project Mariner staff over to its Gemini Agent product.
“A Google spokesperson confirmed the changes, but said the computer use capabilities developed under Project Mariner will be incorporated into the company’s agent strategy moving forward. Google has already folded some of these capabilities into other agent products, including the recently launched Gemini Agent, the spokesperson added.
The change comes as Google and other AI labs rush to respond to the rise of highly capable agents like OpenClaw.”
Anthropic is already ahead of Google with its announcement of Claude Cowork, a desktop interface for interacting with AI agents that makes it possible for non-coders to take advantage of AI agents.
Anthropic describes Cowork’s capabilities and purpose:
“Unlike Chat, Cowork lets Claude complete work on its own. Describe the outcome and cadence, and it takes action and keeps you informed. Come back to the result.
Claude delivers finished work instead of step-by-step updates: a formatted spreadsheet, a memo, a briefing doc. You review, refine, and decide what’s next.
Tell Claude what you want from your desktop or phone. Claude picks the fastest path: a connector for Slack, Chrome for web research, or your screen to open apps when there’s no direct integration.”
Cowork is currently available for download for macOS and Windows.
The boom in agentic AI coding has sent shockwaves through the software publishing industry based on fears that AI coding will make it easier for users to roll their own software solutions. Adobe Inc.’s stock has already lost 33% of its value over the past six months, as have many other software companies.
See also: Complete Crawler List For AI User-Agents [Dec 2025]
Screenshot Of Google Search For Adobe Inc Stock Price
For example, Mistral recently released Voxtral TTS, an inexpensive text-to-speech AI that can run on a laptop with at least 3GB of RAM, undercutting other companies offering the same service for a monthly subscription.
3gb of RAM is wild. been paying ElevenLabs $22/mo for my projects and this runs on hardware i already own. the economics of AI services are going to collapse faster than anyone’s modeling
— Niko Starus (@NikoStarus) March 28, 2026
Google-Agent Connection
The new Google-Agent crawler is labeled as a user-initiated crawler, meaning that the crawler is initiated by a user. The documentation for the new crawler explains:
“Google-Agent is used by agents hosted on Google infrastructure to navigate the web and perform actions upon user request (for example, Project Mariner). It uses IP ranges from user-triggered-agents.json.”
Google currently offers Gemini CLI, but it’s not a one-to-one competitor with the agent-first Claude Code, which is designed to take actions. This addition of the new Google-Agent crawler could be a small piece of a new product that is able to compete more directly with Code.
That said, Google once again finds itself racing to catch up to rapidly developing situations, and this change to the list of Google’s user-triggered fetchers is likely part of Google’s pivot to compete more robustly in the LAM space.
Featured Image by Shutterstock/PPstock
