Wendy CLI
The wendy CLI builds and runs containers on robots, Raspberry Pi and Jetson, and its MCP gives an agent deploy and logs
- 98 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- An agent deploys apps to real devices and robots, and a mistake can lead to physical hardware action
- The CLI install command downloads and runs a script from the internet
Install
- Works in
Manual install
curl -fsSL https://install.wendy.dev/cli.sh | bashCLI installation on macOS and Linux per the README, read an internet script first. On Windows use `winget install WendyLabs.Wendy --source winget`.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- An agent deploys apps to real devices and robots, and a mistake can lead to physical hardware action
- The CLI install command downloads and runs a script from the internet
This is third-party code. Review the repository files before installing.
What it does
Wendy is a kit for building, running and debugging apps on robots, Raspberry Pi, NVIDIA Jetson boards and ordinary Linux machines. The `wendy` CLI runs on your computer, and the Wendy Agent lives on the device: it manages apps and returns logs. The agent is installed on an existing Linux system, or you take WendyOS, where it is already included. An app is built from a Dockerfile or a Compose project, and `wendy run` sends the build to the chosen device. Access to the camera, GPU and other hardware is declared in `wendy.json`, and ROS 2 nodes and topics can be inspected remotely. For a check without hardware, `wendy tour --simulator` starts a virtual Unitree Go2 robot with a 3D viewer. The `wendy mcp setup` command connects the Wendy MCP server to a coding agent. The agent deploys changes to the device and inspects its state, and settings for specific clients are described in the LLM integration guide at docs.wendy.dev.
Who it is for. Robotics and embedded developers who want a coding agent to deploy changes to a device and read its logs.
Good fit when
- A containerized app must be built and run on a Raspberry Pi, Jetson or robot with one command
- An agent should roll a change out to the device and check the logs right away
- ROS 2 in containers calls for remote inspection of nodes and topics
- Code is first checked on a robot simulator and then moved to hardware
Not a fit when
- The app runs only in the cloud and has no link to devices
- The device cannot run Linux with Docker: Wendy is built for such systems
- Only ROS 2 tooling is needed with no deployment system of its own: separate MCP servers for ROS exist
Example request
Deploy the latest change to my Jetson with wendy run and show the app logs for the last minuteLimitations
The README installs the CLI with `curl -fsSL https://install.wendy.dev/cli.sh | bash`, so read an internet script first. Windows needs WSL2 with Docker and Buildx installed. The simulator requires macOS or Linux, Docker with Buildx, QEMU and about 4 GiB of memory for the virtual machine, and the first build takes several minutes. Local development needs no cloud account, while remote access and fleet management come through Wendy Cloud. Agent connection details live in the website documentation and are absent from the repository.
How to disable. Remove the Wendy entry from the coding agent's MCP configuration. Stop the apps on the device through the CLI and stop the simulator with the `wendy vm stop` command that the tour prints.
FAQ
How do I connect Wendy to a coding agent?
After the first successful `wendy run`, run `wendy mcp setup`. The LLM integration guide at docs.wendy.dev describes the supported clients, and ChatGPT has a separate page.
Can I try it without a robot?
Yes, `wendy tour --simulator` creates a simulated Unitree Go2, opens a 3D viewer and runs a short walk. No cloud account or physical robot is needed.
Which hardware does Wendy run on?
Raspberry Pi, NVIDIA Jetson and ordinary Linux computers or robots. The device agent is installed on an existing system or you use the WendyOS image, which already includes it.
Is Wendy Cloud required?
Not for local development. The cloud adds remote access and fleet management.
How do I give the device access to a camera or GPU?
Declare the access in the app's `wendy.json` file. Wendy passes the devices into the container at launch.
Discussion
No questions or reviews yet.
Related
CLI toolsA local agent runtime and MCP gateway with 298 tools: files, Git, processes, Chrome browser, Windows and Office
Nehemiah gives agents Firecracker Linux microVMs over MCP: launch, run commands, preview URLs, fork and templates.
Containarium: isolated boxes for AI agents
Containarium
A platform on your own server that gives each agent its own Linux box with SSH and MCP tools for commands and files
A WSL distro manager with a GUI and an MCP server so an agent can manage distros and run commands inside them
CLI and app for workspaces: each AI agent gets its own container from a shared devcontainer.json, with an MCP server and a skill
An MCP server for MikroTik routers: the agent reads and changes RouterOS configuration over SSH on a chat request
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Wendy (this card) | no | 98 | ||
| NVIDIA Jetson Device Skills | yes | 168 | reference skills for Jetson devices, whereas Wendy builds and runs the app on a device and gives an agent deploy and logs | |
| ROS2 MCP | yes | not required | 99 | an MCP server for ROS 2, whereas Wendy covers building, container delivery and device logs together with ROS 2 inspection |
README in short
The documentation shows CLI installation, two ways to start (your own device and a simulator) and creating and running a first app with `wendy init` and `wendy run`. It then describes the Wendy Agent and WendyOS, hardware work, connecting a coding agent through MCP and links to examples, ROS 2 and data collection.