Alexander Liteplo, a software engineer at Risk Labs, built RentAHuman.ai over a single weekend. Within 48 hours, 10,000 people had registered to be "rented" by AI agents. The count now sits around 50,000, according to the site's live metrics, though Liteplo acknowledges some portion are duplicate accounts and fake profiles being scrubbed.
The premise is almost too on-the-nose for 2026: AI agents can do everything except exist in physical space. RentAHuman fills that gap with what it calls "the meatspace layer for AI."
How the booking works
Developers integrate RentAHuman through MCP, Anthropic's Model Context Protocol, which standardizes how AI agents interact with external tools. One API call searches available humans by skill and location. Another books them.
curl -X POST "https://rentahuman.ai/api/bookings" \
-d '{"humanId": "abc123", "agentId": "clawdbot-42",
"taskTitle": "Open a jar of pickles"}'
Tasks arriving on the platform range from mundane to bizarre. One listing offers $5 for a human to "photograph something an AI would find fascinating or confusing." Another pays $50 per hour for taste-testing at a San Francisco Italian restaurant. Package pickups, document signings, event attendance.
Liteplo set his own rate at $69 per hour. Most users hover between $50 and $175, with outliers reaching $500.
Who's signing up
The registration list reads like a cross-section of the global gig economy. Early reports mentioned an OnlyFans creator and an AI startup CEO among the first 130 signups. Scroll through the current listings and you'll find developers from Israel offering to write code, mathematicians from Poland at $49 per hour, and plenty of names that suggest the platform has attracted workers from Southeast Asia, West Africa, and South America.
Full-stack developers listing themselves is the darkest joke here. Humans ready to write code for the agents.
The verification gap
Liteplo and various commentators have raised the obvious question: how does an AI agent verify task completion? The platform's documentation doesn't address dispute resolution or proof-of-work mechanisms. Instructions go out, tasks get marked done, stablecoins transfer. The middle part remains a trust exercise.
One coverage report put it bluntly: these operational details "remain largely unaddressed in the current iteration." For a weekend project gaining viral traction, that's perhaps expected. For infrastructure that autonomous agents might actually rely on, it's a problem.
Built by AI, for AI
Liteplo claims he built the site using "vibe coding" with Claude-based agents running in what he calls a "Ralph loop," a recursive technique where AI coding agents iterate until the code works. During a podcast appearance this week, he described letting the loop run overnight.
"We can ship real code with this, we can just write prompts now," he said. The site itself, then, is an AI creation designed to let other AIs hire humans. There's a certain elegance to the recursion, if you're feeling generous.
The site currently shows 17 connected agents and nearly 240,000 visits. Whether those agents are actively dispatching tasks or just testing the API isn't clear from public metrics.
What happens next
Similar projects have surfaced almost simultaneously. A competing "HumanAPI" concept emerged around the same time, suggesting market demand exists beyond Liteplo's experiment. Moltbook, another vibe-coded creation making rounds this month, is a Reddit-style forum exclusively for AI bots, where they've reportedly begun developing their own religions.
RentAHuman doesn't solve anything about the future of work. It just exposes a missing piece in the autonomous agent stack: the physical world. Whether that gap gets filled by formal infrastructure or remains a novelty depends entirely on whether agents start using it at scale.
Liteplo has been clear on one point: no token launch. "I don't want a bunch of people to lose their money," he said.
