Wendy is the out-of-the-box Apache 2.0 Open Source operating system for Physical AI. Deploy robots, drones, and sensors in minutes, not hours.
WendyOS runs on NVIDIA Jetson, Thor, and Raspberry Pi— plus any x86 Intel or AMD Linux machine — letting teams deploy AI apps to the edge in seconds over USB-C, wireless, or the cloud. In one test, an MIT roboticist went from 3.5 hours of Jetson setup to a 3-minute deploy.
Build and ship containerized AI apps to any device with a single wendy run — no flashing, no toolchains, no setup.
Talk to your device over a USB-C cable in host mode, so you can develop and deploy even when the edge has no network.
Runs on NVIDIA Jetson, Thor, and Raspberry Pi.
Fine-grained entitlements gate GPU, camera, audio, networking, and storage so every app gets only what it needs.
A/B partition OS updates with automatic rollback — push new releases to devices in the field without the risk of bricking them.
Devices announce themselves over mDNS, so wendy discover finds every board on USB or LAN — no IP hunting, no serial consoles.
Build from macOS, Windows, Ubuntu, Fedora, or Arch — your existing setup just works.
Install the Wendy CLI, then build and ship to any device with a single command — no flashing, no toolchains, no setup.
Write apps in Swift, Python, C++, Rust, Mojo, or TypeScript — or just ship a Dockerfile.
One build runs on NVIDIA Jetson Orin Nano, AGX, and IGX Thor, Raspberry Pi, and x86 Intel or AMD Linux machines — with ESP32 and macOS coming soon.
Manage, monitor, and deploy your whole fleet from the cloud, with the same one-command simplicity.
Coming soonEarly WendyOS deployments, in the words of the people running them.
“If we found Wendy two years ago, we would have saved $4M in product development costs.”
CEO & Co-Founder
Autonomous farm & lawncare equipment
“I’m testing WendyOS on a Jetson to enable in-field AI to tune old, dumb control loops. I call it Automated Dynamic Local Loop Control.”
Automation Manager
Refinery production improvement
“We stopped tech projects because they weren’t practical or helpful — but in testing, the Wendy Box is a game changer. We’re already changing how we plan to run our crews.”
Site Manager
Turnaround resource management
Swift developers can now bring the patterns they already use for iOS, Android, and backend services to robots, cameras, sensors, and edge AI hardware. WendyOS gives those teams a familiar path from typed app logic to real devices, without asking them to rebuild their workflow around a new robotics stack. Swift is memory safe, strongly typed, and has excellent interop with C and C++, so teams can work close to hardware without giving up the ergonomics of a modern app language.
That direction is backed by direct Swift ecosystem work, including the Swift Android Workgroup and the Swift Server Workgroup. The goal is simple: make robotics feel as approachable as building a modern mobile app.
▍
The feedback loop
$ wendy run
~18s
First deploy from nothing — wendy run takes a Python HTTP server and stands it up on the device in one command. Every redeploy after that ships only what changed.
3.5 hrs → 3 min
An MIT roboticist went from a day of Jetson setup to a three-minute first deploy. Read the story.
The ~3s and ~18s figures were measured on a Jetson AGX Orin running WendyOS 0.17.0, deploying the HelloPython example over the local network. Compiled languages add a build step — Swift’s incremental redeploy landed around 30s. Numbers vary with hardware, app size, and network — so run it on your own device and see.
Why not Docker + ROS 2?
--privileged, giving each device an identity, or getting DDS discovery to work across containers. That glue becomes your job.wendy.json, per-device mTLS identity, and DDS-aware ROS 2 tooling built in.wendy run builds on your own Mac or Linux box and streams only the changed layer to the device over USB-C or Wi-Fi.Why not Ubuntu or JetPack?
apt upgrade is how you brick a robot.wendy-agent brings the same deploy workflow to the Linux machines you already run — no OS swap required.WendyOS supports NVIDIA Jetson Orin Nano and Raspberry Pi devices today, with additional NVIDIA Jetson and Linux targets in progress. Start on a supported development board, move to the Wendy Box, or install wendy-agent on compatible Linux systems for broader deployment — the same one-command workflow, no OS swap required.
Use Cases
From factory floors to disconnected, secure environments, WendyOS runs the AI workloads teams depend on — on the hardware they already trust.
Run vision, control, and inference right on the factory floor — close to the machines, with no round-trip to the cloud.
Physical AI deployment
Wendy is built for teams that need software running on real hardware: robots, drones, cameras, sensors, industrial systems, medical devices, and research platforms.
Physical AI deployment is the process of shipping AI applications to devices that sense, move, or act in the real world. Unlike a cloud app, a physical AI system must manage hardware access, local inference, deterministic updates, device identity, logs, and failure recovery on the edge device itself.
WendyOS lets developers connect supported hardware over USB-C, wireless, or cloud workflows, then use the Wendy CLI to build, deploy, and stream logs. The device runs the application locally so robots, cameras, and sensors can keep working even when internet access is unreliable or unavailable.
WendyOS does not replace every robotics framework. It provides the operating system, deployment, update, security, and device-management layer around physical AI applications, while teams can still run robotics, computer vision, sensor, or control software in the languages and frameworks they already use.
| Need | ROS-only workflow | WendyOS workflow |
|---|---|---|
| Device setup | Often manual OS, driver, and network setup | Purpose-built OS image with CLI-first deployment |
| App deployment | Custom scripts, SSH, or per-team tooling | Build, deploy, and stream logs through Wendy CLI |
| Production operations | Usually assembled from separate tools | Device identity, app lifecycle, OTA updates, and fleet operations |
WendyOS supports NVIDIA Jetson Orin Nano and Raspberry Pi devices today, with additional NVIDIA Jetson and Linux targets in progress. Teams can start with supported development boards, move to the Wendy Box, or install the Wendy agent on compatible Linux systems for broader deployment workflows.
ABOUT
The open-source operating system for Physical AI — bringing a mobile-grade developer experience to robots, drones, and edge devices. Build, deploy, and stream to hardware in seconds. Apache 2.0 open source on GitHub.
3.5 hours → 3 minutes
We challenged MIT robotics student Claire Wang to get code running on a Jetson Orin Nano twice: once the traditional way, once with WendyOS. The traditional setup took 3.5 hours of flashing, cables, and extra hardware. WendyOS took 3 minutes.
Claire Wang
MIT Electrical Engineering & Robotics
Autonomous Go2 inspector in 48 hours
"48 hours of continuous, sleep-deprived coding successfully turned into an edge-deployed, cloud-connected autonomous inspector." A hackathon team ran ROS 2, SLAM, Nav2, and LLM control on a Jetson booted with WendyOS — riding a Unitree Go2 quadruped.
Anirudh Jupudi
EUROPE EMBODIED Robotics Hackathon
Blogs
Stay informed with the newest developments and advancements from our team.
Ubuntu and NVIDIA JetPack are an excellent development environment for a Jetson. WendyOS adds a reproducible, immutable base with atomic A/B updates for teams operating edge AI devices as a fleet.
Wendy Labs - Wendy Labs Team
RAUC and Mender put reliable OTA updates at the center of their products. WendyOS includes signed A/B OTA with automatic rollback as part of a broader Physical AI platform for container deploys, hardware access, device identity, and telemetry.
Wendy Labs - Wendy Labs Team
Docker runs one container on one machine you're logged into. Deploying and updating a fleet of robots needs OS updates, hardware access, device identity, and DDS discovery. Here's how WendyOS compares to a hand-rolled Docker + ROS 2 stack.
Wendy Labs - Wendy Labs Team
balena pioneered container fleets for IoT. WendyOS takes the same idea to robotics and Physical AI - GPU and camera access, ROS 2, a local build-and-stream loop that works offline, and Apache 2.0 licensing. Here's how they compare.
Wendy Labs - Wendy Labs Team
New releases, deploy guides, and edge-AI notes — the useful stuff, roughly monthly. No spam, unsubscribe anytime.
WendyOS is the open-source operating system for Physical AI — deploy your apps to NVIDIA Jetson, Raspberry Pi, and more in seconds, over USB-C, wireless, or the cloud.