wendylabsinc/WendyOS
wendylabsinc/WendyOS has 103 GitHub stars and 10 forks. GitHot measured +2 stars in 24 hours, +7 in 7 days, and +8 in 15 days. Last checked 2026-11-17 03:07 UTC. Last push 2026-11-15.
Measured star growth
- in 24h
- +2
- in 7 days
- +7
- in 15 days
- +8
- in 1 month
- Collecting history
Net star changes over overlapping periods, not daily totals. Missing measurements are left blank. Stars measure attention, not software quality.
About
Build, deploy, and debug robot apps with your coding agent. Supports Jetson, Raspberry Pi, Linux, and simulation.
Developer ToolsREADME.md
OriginalWendy
Build and run apps on robots, Raspberry Pis, Jetsons, and other Linux devices. Wendy builds your code, deploys it, and brings the logs back to your terminal. Start with a simulated robot on your own computer, then connect hardware when it is ready.
Documentation · Discord · Examples
What Wendy does
- Simulation: run WendyOS in a local VM with simulated Unitree Go2 and G1, SO-101 and ROSMASTER R2 robots, then deploy the same app to hardware.
- Supported hardware: NVIDIA Jetson Orin Nano, AGX Orin and AGX Thor; Raspberry Pi 3, 4 and 5; Qualcomm Dragonwing; Arduino UNO Q; any ARM64 or x86-64 Linux computer; Apple Silicon Macs (beta).
- Robots: install the Agent on a Unitree G1's onboard computer, and use a Unitree Go2's ROS 2 topics and front camera from your apps.
- AI coding agents: an MCP server and plugins for Claude Code and Codex, a hosted ChatGPT plugin,
wendy chatin the terminal, and persistent agents that talk A2A. - ROS 2: standard ROS 2 in containers,
wendy device ros2to inspect a remote robot, and a Foxglove bridge. - ESP32 (beta): flash Wendy Lite firmware, then build and deploy native ESP-IDF apps with
wendy run. - Wendy Net: connect nearby devices over Ethernet, Wi-Fi, Bluetooth or Wi-Fi Aware so their apps can find each other.
- SensorLink: view a Wendy Lite board's camera from your computer, or pair sensor sources with a device using
wendy device pair. - Developer loop: Stagefile builds that upload only what changed,
wendy run --watch, a shared remote build host, and multi-service apps. - Install and flashing:
wendy installflashes Jetson, Raspberry Pi, Dragonwing, Arduino UNO Q and ESP32 boards from your computer. - WendyOS: a Yocto-based Linux distribution with A/B OS updates through
wendyos-update, automatic rollback, and stable and nightly channels. - Cameras: USB, CSI and IP (RTSP/ONVIF) cameras and ROS 2 image topics all appear to apps as
/dev/videodevices. - Wendy Data: record sensor data, app events and model results around trigger events, and retrieve them as episodes.
- Templates and tutorials: start from
wendy inittemplates in Python, TypeScript, Rust, C++, Swift and Mojo. - Wendy Cloud: enroll devices and reach them and their services through tunnels, without opening inbound ports.
- Security: mutual TLS for provisioned devices, certificate-based enrollment, apps sandboxed by entitlements, and a published threat model.
Start with a moving robot
The guided tour creates a simulated Unitree Go2, opens its 3D view, and runs a short walking demonstration. You can try it without a physical robot or a Wendy Cloud account.
You need macOS or Linux, a running Docker installation with Buildx, and internet access for the initial downloads and build. The robot VM uses four virtual CPUs and 4 GiB of memory; allow disk space for the OS image and containers. Apple Silicon is the fastest path. x86 computers emulate the ARM64 virtual device and can take longer. The first run downloads an OS image and builds the robot software, so allow several minutes. The tour checks Docker and QEMU before starting. See the prerequisite details below if either is missing. Python runs inside the container; you do not need to install it on your computer.
1. Install the CLI
On macOS or Linux:
curl -fsSL https://install.wendy.dev/cli.sh | bash
wendy --version
The CLI runs on the computer where you write code. It does not replace that computer's operating system.
The installer starts the tour when an interactive terminal is available. When
it leaves setup for later, it prints wendy tour and wendy tour --simulator
as next steps.
On Windows, install with winget install WendyLabs.Wendy --source winget.
The managed robot VM tour requires macOS or Linux. Windows users can follow the
physical-device guides or try
the local Docker robot demo.
Other installation methods include Homebrew, .deb, .rpm, and AUR.
Already installed? Run wendy update --check to check your version and see the
right update method. Run wendy update to apply a supported update. Restart your
AI app or Wendy MCP server afterward so it uses the new CLI.
2. Take the tour
wendy tour
Choose the recommended simulated robot path. The tour checks the environment, creates the virtual device, opens its browser viewer, and starts a five-second walk. Look for the Go2 moving in the scene. The tour checks that its position changed and stops the movement before reporting success.
The tour saves the movement settings in ~/Documents/wendy-robot-tour/tour.json.
Change turn_rps to 0 to walk straight instead of turning, then run it again:
wendy tour --simulator
This reuses the same virtual device. The VM and viewer stay available after the
check. Use the exact wendy vm stop command printed in your terminal when you
are finished. To keep another tour project, use wendy tour --simulator --project DIR.
If setup fails, follow the specific prerequisite or connection instruction in the terminal and rerun the tour. See first-deployment troubleshooting for builder, network, or readiness errors.
Install or check the simulator prerequisites
Start Docker, then check:
docker info
docker buildx version
Both commands should succeed. The virtual device also needs QEMU and its ARM64 firmware. On macOS, Wendy offers to install QEMU through Homebrew when needed. On Debian or Ubuntu, install it before the tour:
sudo apt install qemu-system-arm qemu-efi-aarch64
See virtual-device setup for other setup details.
Use your own device
Wendy has two parts: the CLI on your development computer, and the Agent on the target running your apps. WendyOS includes the Agent. Existing Linux installations can keep their OS and install the Agent separately.
| Your target | Setup |
|---|---|
| NVIDIA Jetson Orin Nano or AGX Orin developer kit | Full USB recovery updates boot firmware and OS together. |
| NVIDIA Jetson AGX Thor | USB recovery guide |
| Raspberry Pi 3, 4, or 5 | Write a microSD card and connect the Pi |
| Existing Linux computer or robot | Install Wendy Agent and keep your OS. |
| Unitree G1 | Install the Agent on the development computer and keep the vendor OS. |
| Apple Silicon Mac | Native Mac target, currently beta. |
| ESP32 | Wendy Lite setup, currently beta. |
The board guides list storage, power, and cable requirements before flashing.
Flashing erases the selected target storage. Confirm the intended device boots
and appears in wendy discover before deploying your first app.
For guided device setup, run wendy tour --hardware. Once connected, create a small HTTP app:
wendy init hello-wendy --target wendyos --language node --template simple-api --assistant skip --git-init no
cd hello-wendy
wendy run
Select your device. Open the address Wendy prints and check for
{"message":"hello-world"}. Edit src/index.ts, save, then run wendy run again
to rebuild and deploy your change. Node.js runs inside the container.
Build with a coding agent
After your first successful run, you can let a coding agent deploy changes and inspect the device through Wendy's MCP server:
wendy mcp setup
Follow the coding-agent setup guide for supported tools. For ChatGPT, see the ChatGPT guide.
Keep building
- Tutorials cover Python, Swift, Rust, C++, TypeScript, and more.
- Hardware access covers cameras, audio, Bluetooth, and GPUs.
- Examples include robot control, local AI, and web apps.
- Wendy Cloud adds remote access and fleet management when you need it. Local development does not require a cloud account.