The Wendy ecosystem
Choose the Wendy tools for building apps, connecting sensors, operating devices, collecting data, and analyzing results.
Wendy provides tools to build and operate AI applications on robots and edge devices. Start with the part that matches your task. The products have separate setup requirements and release status; using one does not automatically configure the others.
Choose a starting point
| Your task | Wendy component | Start here |
|---|---|---|
| Build and deploy an application | CLI, templates, WendyOS, and Wendy Agent | Run your first app |
| Test without a physical board | WendyOS VMs and robot simulation profiles | Simulator setup and sim-to-real guidance |
| Ask an AI assistant to develop or inspect an app | Wendy Chat, MCP, and coding-agent integrations | Chat and coding-agent setup |
| Reach devices outside your local network | Wendy Cloud | Cloud v2 and Legacy workflows |
| Run an ESP32 application | Wendy Lite | Native ESP-IDF setup |
| Use a small device's camera stream | SensorLink on supported Lite firmware | Lite cameras and SensorLink overview |
| Record sensors, application events, and model results | Wendy Data | Record an episode or deploy a campaign |
| Analyze operating data and build dashboards | Wendy BI | BI capabilities and access |
| Inspect devices and receive alerts on an iPhone | Wendy Companion | App Store and device notifications |
| Capture human pose demonstrations | TeleCollect | Capture and export pose data |
| Evaluate portable fire monitoring | WendyBox | WendyBox |
Understand the boundaries
For local development, the CLI talks to Wendy Agent on your device. Linux targets run containers; supported Mac targets run native applications. Wendy Lite uses a separate firmware workflow. You can deploy locally without Cloud enrollment. Initial tool, image, and model downloads can still require internet access.
For remote operation, Cloud adds device identities, organizational access, and remote connections. Cloud v2 and v1 use separate enrollments, and v2 tunnels to application ports need updated Cloud, CLI, and Agent versions. Use the migration guide before moving an existing fleet.
Wendy Data records episodes on the device. Upload is a separate configuration, and BI needs its own access and data connection. Recording an episode or enrolling a device does not automatically create a BI dashboard, configure historical backtesting, or train a model.
To alert an operator, your application can request a notification through the Agent. Cloud authorizes and stores it, and Companion provides the operator's inbox and push experience. App registration, grants, recipient access, and a compatible mobile build are required.
Build a workflow in stages
Start with a locally running app and verify its output. Add Cloud access when you need to operate remotely. Use Data to record a meaningful event and inspect the resulting episode. Connect the data source required by your BI use case, then verify its query result. Configure notifications separately when someone needs to respond to an event.
Each stage should produce a result you can check. For example, deploying a campaign arms it; the episode manifest confirms what actually recorded. Cloud accepting a notification does not guarantee a phone received the push.
Availability and security
Use the roadmap for release status and each setup guide for target-specific limits. Companion's App Store release and the reusable SDK have separate release tracks. A simulated robot does not establish that a policy is ready for physical hardware.
WendyOS, the CLI, and Wendy Agent are open source. Check the relevant product's terms and deployment options for hosted services. Local operation, local model inference, Cloud storage, and hosted AI calls have different network and data requirements. See Security for implemented controls and work still in progress.