These days, there’s an app for almost every imaginable user and use case, but one thing they all have in common is that they center around one device: the smartphone.
It changes like today Hugging Face10-year-old New York startup known for being an online space to host and use cutting-edge, open-source AI models, agents, and applications the new App Store for Affordable Miniits inexpensive ($299) open source physics robot It debuted in July 2025 (itself a fruit of Hugging Face the acquisition of another startup, Pollen Robotics).
The new Hugging Face Reachy Mini App Store already has a library of over 200 community-created apps, and Reachy Mini owners will be able to download any of them for free to get started (unlike smartphone apps, this store doesn’t yet have a monetization option for app creators).
The Reachy Mini App Store will also offer Reachy Mini owners—about 10,000 units sold since last year—an easy way to create their own custom apps for the small, stationary desktop robot with built-in camera eyes, a speaker, and a microphone. Hugging Face’s existing, AI-powered agent called "ML Intern."
The importance is not only in the hardware, but also in its elimination "roboticist" obstacle; for the first time, people with no engineering or coding experience ship a functional robotics program in less than an hour.
"Anyone can create apps," Clément Delangue, CEO and co-founder of Hugging Face, told VentureBeat in a video interview. "My intuition is that more and more (AI) model builders will be released on the Reachy Mini to test the robotics capabilities of new models."
Make robots as accessible to common people as computers and smartphones
A technical bottleneck in robotics has historically been the scarcity of high-quality training data.
Large Language Models (LLMs) have mastered general-purpose coding by training in massive repositories like Microsoft. GitHuba dedicated code volume remains for robotics "little one" In comparison (though Github probably contains the largest library of bot code available to date. Over 17,000 different repositories or "repos" dedicated to the field).
This lack of information has meant that until now AI agents have been relatively poor at understanding physical abstractions and hardware software requirements.
Hugging Face’s solution is one agent toolkit acts as a mediator. Instead of forcing the user to learn a specific robotics SDK or master the nuances of the robot’s software, the toolkit allows the user to describe any behavior in plain English, e.g. "wave when someone says good morning".
The AI agent then does the heavy lifting: writing the code, testing it against the robot’s specific limitations, and sending the final package.
"Historically, it has been very difficult" Delangue spoke to VentureBeat about creating robotics applications. "But we’ve worked really hard on this topic by open-sourcing everything we do, working on the right abstractions for robotics, and making it easy for agents to understand and use."
The platform is model-agnostic, supporting a wide range of leading intelligence engines. Users can build apps using Hugging Face’s own ML Intern agent or use external models including GPT-5.5, Claude Opus 4.6, Kimmy 2.6, Mini Max GM5 and Deep Sig V4 Pro.
The official chat apps use OpenAI Realtime and Gemini Live for real-time interaction. By providing these high-level abstractions, Hugging Face subverts the traditional "weeks of integration" robotics becomes a process that takes minutes.
The cheap Reachy Mini is a hit
To take advantage of the new Hugging Face Reachy Mini App Store, users are encouraged to purchase Affordable Minicute desktop robot Hugging Face was reintroduced in July 2025 As a cost-effective, open-source alternative to existing, commercial robots from popular robots such as Boston Dynamics. Spot the robot dog sells for about $70,000. Even Chinese competitors start at $1,900.
In contrast, the Reachy Mini is accessible to hobbyists and developers. It comes in two variants:
-
Reachy Mini Lite ($299 plus shipping): A bundled version that connects via USB and uses an external computer for processing.
-
Reachy Mini Wireless ($449 plus shipping): Standalone version equipped with Raspberry Pi CM 4 and Wi-Fi connectivity.
Of the 10,000 Reachy Mini units sold so far, 3,000 have been sold in the past two weeks alone, Delangue said. Hugging Face expects to ship another 1,000 units over the next 30 days.
Even those who don’t own a Reachy Mini can still develop apps for it using the Reachy Mini App Store and the Reachy App, which includes a 3D simulation of the robot and its responses.
The App Store itself is hosted on the Hugging Face Hub. It works like a standard firmware repository, but for hardware behaviors:
-
Search and Install: Users can find apps, click a button and install them directly on their robots.
-
Fork: There is every program "hook," means that the user can duplicate an existing program and ask the AI agent to modify it (e.g. "Answer in French").
-
Simulation mode: The main thing is that the store has a browser-based simulator. This allows users who do not own a physical Reachy Mini to create, test and play with a catalog in a virtual environment.
Both are sequels to Hugging Face "Robot" effort – a project that started in 2024 With Hugging Face researchers specializing in robotics and artificial intelligence, it develops and publishes its own open source code, tutorials, and hardware to make robotics development more accessible to a wider audience.
Unlike Github, which is designed for a developer audience, the Hugging Face Reachy Mini App Store is designed for robot owners and users with no technical experience or training.
We continue with open source ethics and practices
Hugging Face’s strategy relies on closed-source hardware and software "almost impossible" build to scale.
Delangue points out that closed systems hinder the development of agents and limit society’s ability to innovate. As a result, the entire Reachy Mini platform is open source.
This open licensing model has two main implications for the ecosystem:
-
Rapid Development: Because the code is public and integrated with the Hugging Face ecosystem through "gaps," Hugging Face’s feature to host AI-powered web applications Launched in 2021agents can more easily learn how to interact with the hardware.
-
Community Sovereignty: Apps are not locked behind a private wall. Currently, all 200+ apps in the store are free, even though the platform is based on the foundation "Gaps" giving creators the flexibility to potentially monetize their work in the future.
"Currently, all programs are free," Delange noted. "It’s flexible, (Face Hugging) Built on loopholes, so at some point people will pay for them."
Robotics is entering the era of an accessible hobby
Hugging Face’s Reachy Mini App Store launches with 200 apps already available.
Who built them and how did they do it without this platform before?
Delangue told VentureBeat more 150 different creators contributed to the shop, most of whom had never written robotics code before.
However, thanks to Hugging Face’s ML Experience and Github. The new Hugging Face Reachy Mini App Store now brings together tools and existing apps for easy accessibility.
Delang wanted to highlight one of the first Reachy robotics programmers in particular to VentureBeat: Joel Cohen, a 78-year-old retired marketing executive.
Cohen, who is color blind and has no technical knowledge, spent two weeks assembling the Reachy Mini Lite (a task that normally takes three hours). Despite these physical challenges, he used an artificial intelligence agent "VP of Future Thinking" A facilitator for Zoom-based CEO peer groups. The program allows the robot to:
-
Greet 29 members by name.
-
Real-time fact-checking discussions.
-
Summarize key themes and retract surface-level answers.
"I set it up by describing what is needed in plain English," Cohen told VentureBeat in a press release ahead of the presentation. "There is no SDK. No robotics background. No developer experience".
Other community-driven apps include:
-
Emotional Damage Chess: A robot that plays chess and mocks the user’s blunders.
-
Reachy Phone House: An anti-procrastination tool that prompts the user to pick up their phone and get back to work.
-
Language Tutor: A physical companion that listens to speech and corrects accents.
-
F1 Race Commentator: A tablemate who calls Formula 1 races live as they happen.
Delangue himself told VentureBeat that he programmed his Reachy Mini robot to act as a receptionist in just a few hours at the Hugging Face Miami office.
“It basically does facial recognition to detect when you’ve come into the office, and then it looks at you and hires you." Related to Delang. "It says, ‘Hey, welcome to the office. Who are you here to see?’ Then he sends me a message: ‘Karl just arrived at the office. He’s here to meet you, and for these reasons.”It works a bit like my welcome booth at the office, and it took me less than two hours to set up.”
Even for an experienced founder and developer like Delangue, building apps for a robot was out of the question until the combination of Reachy Mini and ML Intern.
“It would be impossible for me," CEO of Hugging Face said. "If you weren’t a robotics developer, it would probably be impossible or take months."
The democratization of robotics
The launch of the Agency App Store marks a fundamental shift in the way we interact with machines. For sixty years, the field has been in demand for deep technical expertise.
By combining low-cost outdoor hardware with the reasoning capabilities of modern artificial intelligence agents, Hugging Face moves toward a future where hardware is a commodity and behavior is limited to what the user can describe.
As Delangue mentioned during the presentation, the goal was to provide a platform for people "I want to get into robotics but don’t have the equipment or skills".
Now with about 10,000 robots "in the wild" and Reachy Mini, a thriving store of agent-written applications, has become the most widely distributed open-source desktop robot in history.
The question is no longer how to build the robot, but what we will ask them to do now that the door is open.
