Google-Agent Could Mark OpenClaw Pivot

Google has quietly updated its list of user-triggered crawlers to include a new entry called Google-Agent. This tool is tied to the company’s Project Mariner, which originally functioned as a browser-based AI assistant, and it may now signal a broader strategic shift. Analysts suggest this could be Google’s way of competing with the emerging popularity of OpenClaw-style personal AI agents.

The Rise of OpenClaw

OpenClaw represents a new generation of personal AI assistants capable of handling a wide range of tasks online. Its standout feature is the ability to form “teams” of specialised agents, with a single orchestrator assigning tasks across the network. These agents can operate from laptops, desktops, and hosted cloud environments. They are model-agnostic, allowing them to connect to providers such as Anthropic (Claude), Google (Gemini), and OpenAI.

The appeal of personal AI agents is growing rapidly, not just in the US and Europe, but also in Asia. Chinese AI providers, including MiniMax, Moonshot AI (Kimi), Alibaba Cloud (Qwen), and DeepSeek, have become increasingly popular because their services are significantly cheaper than mainstream US providers. This has further fuelled the boom in personal AI agent adoption.

The market has become so competitive that OpenAI recently hired Peter Steinberger, the developer of OpenClaw, highlighting the growing importance of personal AI agents in the broader AI ecosystem.

Google’s Project Mariner

Project Mariner was first announced in 2025 and was initially available only to Google AI Ultra subscribers and select Google Labs testers. The tool was designed as a browser-based assistant that could navigate the web and complete tasks on a user’s instruction. Early demonstrations, however, suggested that it was somewhat clunky, with one tester describing the interface as “far from perfect.”

Despite its limitations, Project Mariner provided Google with a foundation for developing autonomous AI capabilities, allowing the company to experiment with agent-based technology that could take actions beyond just generating text.

Large Action Models (LAMs)

The AI agent trend has evolved into a broader phenomenon known as Large Action Models (LAMs). Unlike traditional language models (LLMs) that primarily generate text, LAMs can understand a user’s objective, break it down into steps, interact with APIs, click buttons, and complete tasks either autonomously or under human oversight.

LAMs are increasingly used in areas such as software development, WordPress plugin creation, content generation, and social media management. They are effectively autonomous “robot workers” capable of executing real-world tasks without constant supervision.

The release of WordPress 7.0, designed to be more AI-friendly, is expected to accelerate the adoption of LAMs, giving businesses new ways to manage websites and automate workflows.

Transition from Mariner to Gemini Agent

Many of the Project Mariner staff have been reassigned to Gemini Agent, with insights and technology from Mariner being integrated into Google’s broader AI agent offerings. A spokesperson for Google confirmed that the capabilities developed under Mariner will now form part of the company’s overall agent strategy, including integration into Gemini Agent.

This move appears to be part of a response to competitors like OpenClaw, whose sophisticated agent capabilities have set new benchmarks for personal AI. Google is clearly aiming to catch up and strengthen its position in the growing AI agent market.

Competitor Innovations

Anthropic has taken a lead with its Claude Cowork desktop interface, which allows non-coders to use AI agents effectively. Unlike standard AI chat interfaces, Cowork can complete tasks autonomously, delivering finished outputs such as spreadsheets, reports, or memos. Users provide outcomes and timelines, and the AI determines the most efficient path to complete the work, whether that involves integrating with Slack, Chrome, or other tools. Cowork is currently available on both macOS and Windows.

The rapid development of agentic AI has also impacted the software industry. Stocks of major software companies, such as Adobe Inc., have fallen sharply in response to fears that AI agents will make it easier for users to create custom software solutions independently. Over the past six months, Adobe has lost around 33% of its value, reflecting wider industry disruption.

Affordable AI tools like Mistral’s Voxtral TTS, which runs on laptops with as little as 3GB of RAM, are further undercutting subscription-based competitors, demonstrating how accessible AI agent technology is becoming.

Google-Agent’s Role

The newly added Google-Agent crawler is user-initiated, meaning it acts when triggered by a user. It is designed to support agents hosted on Google infrastructure, enabling them to navigate the web and complete tasks on demand. Project Mariner is an example of one such agent that could utilise this system.

Although Google currently offers Gemini CLI, it is not yet a direct competitor to agent-first solutions like Claude Code, which are built for autonomous action. The addition of Google-Agent could be a small but crucial step in Google’s broader strategy to develop AI agents that can actively perform tasks rather than just generate responses.

The Competitive Race

Google is clearly moving quickly to keep pace with the evolving AI landscape. With the rise of LAMs and highly capable personal AI agents like OpenClaw, the company’s pivot reflects the urgency of maintaining competitiveness in a market that is expanding rapidly. The Google-Agent update may be a strategic attempt to ensure its AI tools remain relevant as the capabilities and expectations of users continue to grow.

Conclusion

The launch of Google-Agent marks a subtle but significant shift in Google’s approach to AI. By leveraging the lessons from Project Mariner and integrating them into Gemini Agent, Google is positioning itself to compete in the fast-growing space of Large Action Models. Meanwhile, competitors such as OpenClaw and Claude Cowork are setting high standards for what personal AI agents can achieve. The coming months will reveal how these tools reshape both the AI agent market and broader technology trends.