Humanoid Robot App Stores, Compared

  • Early-stage category — tracked and updated as it develops

Last updated: August 2026

The Robot App Stores

Two real platforms exist today. Both let a robot owner install new “skills” — packaged behaviors
— without writing code.
PLATFORM LAUNCHED BUILT BY WORKS WITH CATALOG MODEL
UniStore May 7,
2026 		Unitree Unitree G1 only
single-brand 		24 skills at launch Closed platform, open SDK for submissions
OpenMind Jan 29,
2026 		OpenMind
(independent) 		UBTech, AgiBot, Deep Robotics, Fourier, Booster, Dobot, LimX, Magic Lab
cross-brand 		1,000+ developers building; catalog still forming Open OS (OM1), open to any manufacturer

Platform Profiles

A closer look at each platform’s approach — and what it signals about their strategy.

UniStore

Unitree Robotics · Hangzhou, China
Unitree’s own app store, built exclusively for its G1 humanoid. Owners browse an action library from a phone app, tap to install, and the robot receives the skill over the air. No coding required on the user side.
  • Launch catalog: Boxing drills, Charleston and Michael Jackson–style dance moves, martial arts postures, striding and obstacle-stepping locomotion

  • Strategy signal: Software-first play from a hardware-first company — Unitree is trying to win the way a phone maker wins with an app store, not just with specs

  • Open questions: Pricing, revenue share for developers, safety review process, and skill portability to Unitree’s other robots (H1, R1, Go2) are all undisclosed
Single manufacturer
Entertainment-first catalog

OpenMind

OpenMind · Independent software company
A manufacturer-agnostic app store built on OpenMind’s own operating system, OM1. Instead of locking skills to one robot brand, OpenMind wants a single skill to run across many — the explicit goal is to be the operating layer robots run on, not just a catalog for one company’s hardware
  • Target skill categories: Elder care, personalized education, home cleaning, home security — utility-first rather than novelty-first

  • Strategy signal: Betting that hardware makers won’t want to build their own closed ecosystems and will plug into a shared one instead

  • Open questions: Whether any hardware maker will commit exclusively, and whether cross-brand skills actually perform consistently across different robot bodies
Cross-manufacturer
1,000+ developers

The Layer Underneath: NVIDIA

Neither app store operates in isolation — most of the skills that eventually reach a store were likely built using tools from a company that isn’t in the app-store business at all.

Not a competing store — the infrastructure both stores run on

NVIDIA’s Isaac platform (Isaac Sim, Isaac Lab, and the Isaac GR00T foundation models) is the toolkit developers across the industry use to train and test robot skills in simulation before they ever reach a real machine. It’s closer to Xcode or Android Studio than to the App Store itself — the workshop, not the storefront.
Nearly every major humanoid manufacturer builds on it, regardless of which app store — or none — they eventually ship through.
Boston Dynamics
Figure
Agility Robotics
1X
AGIBOT
Fourier
Skild AI
+ dozens more

Why This Matters

The hardware race gets the headlines — who builds the most capable robot body. But the smartphone era was won in the software layer, not the hardware spec sheet. The open question worth watching isn’t which robot walks best — it’s whether the winning ecosystem turns out to be closed and single-brand (UniStore’s bet) or open and cross-brand (OpenMind’s bet) — and whether NVIDIA’s infrastructure layer ends up mattering more than either.

Track this space as it develops

New platforms, catalog growth, and developer economics — delivered as this category matures.

    Sources include company announcements from Unitree and OpenMind, NVIDIA developer program disclosures, and industry reporting current as of July 2026. This is a nascent, fast-moving category — figures such as catalog size and developer counts will change. See the Humanoid Robot Directory →