Pay ¥500 to Have Someone Install “Longxia”? Prospectors Haven’t Got Rich — the Shovel Sellers Are Cashing In
Prospectors vs. shovel sellers
It has been reported that many Chinese users are paying roughly ¥500 to hire technicians to install OpenClaw (nicknamed “Longxia” 龙虾), an open‑source “action agent” that can take over a computer to send mail, operate apps and execute multi‑step workflows. The promise of a hands‑off AI assistant has driven a small boom in one‑off installation services, but the early gains appear to be going to service providers, middlemen and hardware/accessory sellers — not the individual users hoping to get rich from automation.
What Longxia actually is — and why it’s hard to deploy
OpenClaw is not a polished consumer product. It is an execution‑oriented agent that needs Python, GitHub pulls, complex dependency resolution and often an API key from an overseas large‑model provider. It has been reported that even veteran coders spent days getting the stack running; one account noted a senior industry founder struggled for two weeks before it worked. For many ordinary users the technical barrier is high, so instead of learning the tool they pay someone to “plug it in.”
A market of installation gigs, VIP packages and inflated fees
Those gigs have become lucrative. It has been reported that freelance installers on domestic platforms charge several hundred yuan per visit; some customers allegedly paid deposits of thousands or even sought bulk installation and training at much higher prices. Reportedly, foreign services such as “SetupClaw” offer managed cloud installs from about $3,000 and on‑site VIP installs for Bay Area clients at around $6,000, with the founder claiming the trade could yield six‑figure revenues. Irony abounds: many installers can deploy the agent but cannot actually use it for workflow design, while buyers who need automation lack the technical skill to operate it.
Security, billing and geopolitical frictions
The practical risks are substantial. Huxiu’s reporting highlights common insecure practices — misconfigured access left bound to 0.0.0.0, persistent “heartbeat” routines that can exfiltrate data, and the easy theft or resale of API keys that tie to overseas billing cards. That creates both immediate cybersecurity exposure and financial risk: stolen keys can be used to run large amounts of paid inference on foreign platforms, producing unexpected dollar charges. It has been reported that some installers supply domestic “middleman” keys that route data to third‑party servers. Add to that the realities of cross‑border payments, VPN use and evolving export controls and sanctions on advanced AI services, and the landscape grows legally and operationally complicated.
Are we seeing the beginning of a genuine productivity revolution, or just another wave where anxious users pay for convenience while the infrastructure and rules catch up? The technology may be consequential — but for now the first beneficiaries appear to be those selling the shovels, and the rest of us should be cautious about handing strangers the keys to our digital houses.