Baidu Inc. launched DuClaw this week, a cloud-managed platform that gives users zero-deployment access to OpenClaw AI agents, just as Nvidia CEO Jensen Huang declared that “every company needs to have an OpenClaw strategy.” The twin developments signal that the open-source agent framework, which Nvidia calls the fastest-growing open-source project in history, is now pulling major Chinese and American tech players into a race to define how AI agents reach everyday users. In China, the enthusiasm has taken on a life of its own: hobbyists across the country describe themselves as “raising lobsters,” a nickname for the process of training and customizing OpenClaw agents.
DuClaw Strips Setup Friction From OpenClaw
Most open-source AI tools still demand server provisioning, model configuration, and API plumbing before a developer can test a single agent. DuClaw aims to skip all of that. The platform, managed by Baidu AI Cloud, offers a web interface where users can spin up OpenClaw agents without touching infrastructure. It ships with pre-built Baidu skills, including Baidu Search, Baike, and Scholar, and supports multi-model configurations so agents can draw on different large language models depending on the task.
DuClaw’s positioning reflects Baidu’s broader cloud ambitions. By moving agent deployment into a managed environment, Baidu can bundle compute, orchestration, and observability into a single subscription-style product. For enterprises that have been wary of open-source AI because of integration overhead, a hosted control plane lowers the barrier to experimentation. It also offers Baidu a way to differentiate its cloud from rivals by leaning into the OpenClaw ecosystem rather than proprietary agent stacks.
Planned integrations with WeCom, DingTalk, and Feishu would embed those agents directly inside the workplace messaging apps that dominate Chinese enterprise communication. If Baidu delivers on that roadmap, a sales team on DingTalk or a customer-service desk on WeCom could deploy an always-on AI assistant without writing a line of code. That is a meaningful shift from the current state of agent adoption, where most deployments still require dedicated engineering teams.
DuClaw also underscores how distribution is becoming as important as raw model quality. The ability to launch an agent from a browser, connect it to internal tools, and expose it inside chat interfaces is what turns an open-source framework into something that non-technical staff can actually use. In that sense, DuClaw is less about novel algorithms and more about packaging: it turns OpenClaw from a GitHub project into a service that fits into existing enterprise workflows.
Redfinger Operator Brings Agents to Phones
DuClaw is not Baidu’s only OpenClaw play. Earlier this month, Baidu AI Cloud released an Android app called Redfinger Operator, billed as the first mobile application built around OpenClaw’s “lobster” agents. The app enables cross-app operations on a phone, handling tasks like ride-hailing and food delivery by coordinating actions across multiple installed services. An iOS version is expected but has no confirmed release date.
The mobile angle matters because it moves OpenClaw agents out of browser tabs and cloud dashboards and into the device people carry all day. If an agent can book a car, order lunch, and check a knowledge base without the user switching apps, it starts to function less like a chatbot and more like a personal operating layer. That is the bet Baidu is making, and it aligns with a broader push the company signaled at its Lobster Market event in Beijing, where it also connected the framework to smart speakers as another surface for agentic AI.
By tying OpenClaw to both phones and household devices, Baidu is sketching out a multi-surface future in which agents follow users from the office to the living room. A commuter might ask a smart speaker to schedule errands, then have a Redfinger-powered agent execute those tasks on their phone while they travel. Each new surface increases the number of contexts in which agents can act, which in turn generates more behavioral data to refine those agents over time.
Jensen Huang Frames OpenClaw as “the New Linux”
While Baidu builds consumer and enterprise on-ramps in China, Nvidia is working to position OpenClaw as the default infrastructure layer worldwide. At the Morgan Stanley Technology, Media and Telecom Conference, Huang discussed agents at length and explicitly named OpenClaw, framing the framework around token scale and the idea that agents improve through feedback and training. He followed up at GTC 2026 by calling OpenClaw “the new Linux,” casting it as an operating system for AI agents rather than a single product.
That comparison is strategic. Linux succeeded not because one company controlled it but because it became the shared substrate that thousands of companies built on top of. Huang appears to want the same dynamic for OpenClaw, and Nvidia has a clear financial interest: more agents running at scale means more demand for GPU compute. At GTC, the company ran a “build-a-claw” event where attendees could deploy always-on agents using Nvidia hardware or local machines. The message was direct. The agent era runs on Nvidia silicon.
Huang’s framing also helps legitimize OpenClaw for conservative IT buyers. Comparing the framework to Linux signals that Nvidia sees it as a stable, long-term layer rather than a passing experiment. For developers, the implication is that learning OpenClaw’s abstractions (its way of defining tools, memory, and feedback loops) could be as career-defining as learning Unix once was.
China’s “Lobster-Raising” Culture and Its Risks
OpenClaw’s rapid global uptake has produced a wave of experimentation in China, where Baidu is one of several firms tying their products to the framework. Enthusiasts there describe themselves as “raising lobsters,” a playful reference to the claw-themed branding and the hands-on process of training agents. The phenomenon cuts across demographics. Schoolkids and retirees alike have taken up the hobby, customizing agents for everything from homework help to daily errands.
That grassroots energy is a competitive advantage for any platform that can channel it, but it also introduces friction with regulators. “Its development is still unstoppable, but the security capabilities also need to keep up so in that sense, this may also be an opportunity,” one observer noted in a Reuters interview, capturing both the momentum behind OpenClaw and the pressure to build guardrails. Chinese authorities have already signaled that agentic systems must comply with existing content and data rules, and the more ordinary users customize agents, the harder that oversight becomes.
The “lobster-raising” culture also hints at how people may come to think about AI agents more generally. Instead of static tools, agents become quasi-pets or apprentices that require care, feedback, and occasional correction. That mindset could make users more tolerant of early mistakes, but it also risks normalizing opaque behavior if people treat agents as inscrutable companions rather than software that can and should be audited.
Media, Metrics, and the Battle for Narrative
As Baidu and Nvidia push OpenClaw deeper into mainstream awareness, media infrastructure is racing to keep up. Wire services and corporate newsrooms are competing to frame the agent boom, with platforms like PR Newswire’s media hub and its associated distribution tools helping companies seed their narratives across global outlets. Baidu’s DuClaw launch, for instance, arrived with detailed technical positioning and customer scenarios that were quickly echoed in coverage of the broader OpenClaw frenzy inside China.
Those narratives matter because they shape how policymakers and investors interpret the stakes. If OpenClaw is seen primarily as a developer playground, regulators may focus on code hosting and license terms. If it is framed as critical infrastructure (the “new Linux” for AI), then questions of sovereignty, security, and interoperability come to the fore. Baidu’s emphasis on compliant, cloud-managed agents and Nvidia’s focus on open, hardware-accelerated ecosystems represent two different answers to the same question: who gets to define the rules of the agent era?
What Comes Next for OpenClaw
The convergence of DuClaw’s zero-deployment model, Redfinger’s mobile reach, Nvidia’s infrastructure push, and China’s lobster-raising subculture points to a near future in which AI agents are both ubiquitous and contested. For users, the upside is clear: agents that can book travel, summarize meetings, and coordinate errands without constant micromanagement. For companies, the opportunity lies in owning the platforms that make those agents easy to build and safe to trust.
The unresolved questions revolve around governance and control. As more of daily life is intermediated by agents, small design choices (how feedback is collected, how tools are vetted, how data is stored) will have outsized impact. Baidu’s cloud-managed approach and Nvidia’s open-source evangelism are early attempts to answer those questions at scale. Whether OpenClaw ultimately resembles Linux, Android, or something entirely new will depend less on any single product launch than on how millions of people, from Beijing hobbyists to Fortune 500 CIOs, choose to raise their own lobsters.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
