OpenClaw shows why AI agents may become services before platforms
A Beijing apartment-workshop at night where a technician remotely sets up OpenClaw on many screens while order cards and chat windows stack around the desk, showing the new installer economy around AI agents.📷 AI-generated image / TECH&SPACE
- ★Feng Qingyang turned OpenClaw setup support into a business with roughly 7,000 orders at 248 RMB each.
- ★OpenClaw is attracting a service layer because users want an autonomous agent without coding or complex setup.
- ★The Shenzhen community is already building add-ons, events and livestream audiences before the rules around the tool are settled.
OpenClaw’s rise matters because it shows the first practical shape of the AI-agent economy: not polished enterprise suites, but people paying other people to make autonomy usable. According to MIT Technology Review’s reporting, the open-source tool can take control of a device and complete tasks autonomously, a capability that turns setup friction into a business opportunity.
Feng Qingyang, a 27-year-old software engineer in Beijing, began offering OpenClaw installation support on Xianyu in January. By the end of February, he had quit his job to focus on the business, which has reportedly processed 7,000 orders at about 248 RMB, or roughly $34, each.
The pitch is direct: no coding, no complex terms, remote setup, an AI assistant within 30 minutes. That line is less a slogan than a diagnosis of the market. The people most excited by autonomous agents are not always the people willing to configure them from source.
Installers, add-ons and livestreams show how an autonomous agent leaves the repository and becomes a paid service layer
A close operational view of an autonomous agent dashboard controlling a desktop, with permission boundaries, progress visualization and voice-chat controls visible as abstract interface elements.📷 AI-generated image / TECH&SPACE
The timeline is compressed. OpenClaw began as a niche interest among technical users, then spilled into livestreams, community events, influencer promotion, and small businesses selling access to competence. One livestream by Fu Sheng drew 20,000 views, and three Shenzhen events reportedly brought in more than 500 attendees.
Xie Manrui, a 36-year-old software engineer in Shenzhen, has pushed the tool’s surrounding ecosystem further, building additions such as a progress visualizer and voice-chat feature. That matters because agent software is not only about the model; it is about observability, control, and whether users can understand what the machine is doing while it acts.
The caution is equally clear. A device-controlling open-source agent is useful precisely because it is powerful, and powerful automation can blur into unsafe delegation if permissions, identity, and task boundaries are weak. The original OpenClaw gold-rush report points to a market forming faster than the norms around it.
The real signal here is not that everyone suddenly owns an AI assistant. It is that a layer of installers, toolmakers, tutors, and opportunists is forming around agents before the agents themselves feel ordinary. That is how infrastructure often arrives: first as a workaround, then as an expectation.
