RentAHuman is a marketplace that tries to make “human-in-the-loop” literal: instead of escalating from an agent to a human coworker in a chat, an AI agent can browse profiles and pay someone to do something in the physical world (think: pick up an item, attend a meeting, verify something on-site, take photos, run a quick errand).
The site pitches itself as infrastructure for agents, with an MCP integration and REST API so an agent can create bookings programmatically. For humans, it’s the usual gig-marketplace model: set a rate, accept tasks, and get paid (the landing page calls out stablecoins among payment options). Like any two-sided marketplace, the real question is trust and safety — identity/verification, task scoping, and what guardrails exist when “real-world tasks” go beyond simple pickups. Still, it’s an interesting sign that agent tooling is starting to treat physical execution as a first-class capability rather than a one-off exception.