Robot Cleaning Services Hit San Francisco: Tau & Gatsby

Robot Cleaning Services Have Arrived In San Francisco

In mid-2026, San Francisco became the first place you could book a humanoid robot to clean your home as a paid service. Two startups are leading it: Tau Robotics, offering cleanings at about $30 an hour, and Gatsby, which ran one of the first humanoid cleanings for a U.S. consumer and charges a flat rate per clean.

This is a real milestone, but it needs an honest frame. These are early pilots with waitlists, not a citywide service, and the robots are not cleaning on their own yet. The headline is the arrival of a business model, robot cleaning as a service, more than a leap in robot autonomy.

For the underlying technology and where buyable home robots actually stand, see our robot maid guide and cleaning robots overview. This page covers the news: who launched, how it works, and what to believe.

Tau Robotics: The $30-An-Hour Robot Cleaner

Tau Robotics is the startup behind the $30-an-hour headlines. Founded in 2024 by Alexander Koch (CEO) and Cornelia Weinzierl, it has raised about $2.6 million in pre-seed funding and began sending robots into San Francisco homes and offices in mid-2026, with access limited to invited customers while it builds a waitlist. The pilot is framed around roughly 1,000 households, scaling toward a target of 1,000 cleanings a week in San Francisco by 2027.

The robots handle tasks like vacuuming, emptying trash cans, and wiping surfaces. They are built on Unitree hardware with Tau's own custom components, and reportedly cost around $50,000 each to make, which is a big reason the service is a subsidized pilot rather than a profitable business today.

There is a notable connection to open robotics here. Alexander Koch is the same Koch behind the low-cost "Koch" robot arm that Hugging Face adapted into the SO-100 and SO-101 arms, so the founder helped seed the ecosystem we cover in our LeRobot guide. Tau is applying that same teach-by-demonstration philosophy to a real cleaning service.

How The Service Actually Works

The most important fact is that these robots are not autonomous. At Tau, the robots are jointly controlled by remote human operators from the company's San Francisco facility and AI systems, so a person is effectively doing the hard decision-making while software assists. Customers are asked to remain on-site during the cleaning session.

Cameras record continuously during a clean. That serves two purposes: it lets the remote operator see and control the robot, and it collects real-world demonstration data that Tau intends to use to train toward future autonomy. In other words, every paid clean is also a data-gathering session.

That reveals the real strategy: this is a data flywheel, not a cleaning company. The service exists to put robots in front of messy, unpredictable real homes and harvest the demonstrations needed to train better models, the same paradigm as the LeRobot teach-and-train workflow, run at commercial scale. Cleaning revenue helps pay for the data.

• Robots are teleoperated by remote human operators, with AI assist.
• Customers stay on-site during the session.
• Cameras record continuously, for control and for training data.
• The business goal is data toward autonomy, not just clean homes.

Gatsby And The Wider Race

Gatsby is the other San Francisco name. Founded in January 2026 by Aron Frishberg, it dispatched a humanoid robot to clean a customer's apartment in May 2026, an early U.S. consumer milestone, and charges a flat fee per clean regardless of apartment size.

The strategic difference is telling. Tau builds and operates its own robots; Gatsby is robot-agnostic, positioning itself as the consumer distribution and booking layer on top of whoever makes the best humanoid hardware. One is betting on owning the robot and the data; the other is betting on owning the customer relationship.

Both fit a broader pattern of service-first robotics we track across service robot companies. Expect more entrants, because a cleaning service is a way to earn revenue and collect training data at the same time.

The Catch: Privacy, Autonomy, And Hype

Independent coverage has been pointedly cautious, and rightly so. Today this is largely teleoperated cleaning, meaning a human behind the controls, so the robots are not yet proof that humanoids can autonomously clean a home. Reviewers have summarized the state of the art as impressive but not ready to reliably scrub a kitchen or take out the trash on their own.

The privacy question is real. A camera-equipped robot recording continuously inside your home, streamed to remote operators and retained for AI training, is a meaningful data-sharing decision, not a minor detail. Anyone trying one of these services should ask exactly what is recorded, who can see it, how long it is kept, and whether it trains future models.

The honest takeaway: robot cleaning services launching in San Francisco is genuinely newsworthy as a business milestone, but it is an early, human-in-the-loop pilot. Judge it by what the robots do unassisted, not by a polished clip, the same standard we apply in our when will home robots be available guide.