Get started
Give your robot a task and let it act on what it sees.
RoboAPI sends camera images, robot state, and a task instruction to Poke-Aurora, which returns the next actions. Use the local console to operate your robot, or call the inference API from your own application.
1. Create an API key
Sign in to the Developer Console to create a key and check your available models and balance. See Account and API keys.
2. Choose your integration
Use your own robot
Connect your hardware, install RoboAPI, and send tasks from your browser.
Call the model from code
Send an image, current state, and instruction. Read your first action sequence.
Use your own robot. Run RoboAPI on the Linux computer connected to your hardware. It collects observations, requests actions, and executes them. Open its local console from your personal computer after setting up the SSH forwarding described in the installation guide. Start with hardware, then installation and calibration. RoboAPI manages task sessions and execution receipts automatically.
Call the model from code. Use the OpenAI Python SDK, JavaScript, or HTTP. Your application collects observations and translates the returned actions into commands for its controller. Make your first API call, then read Observations and actions and the API reference.
3. Run a task
After setting up and calibrating your hardware and connecting the model service, enter a concrete task in the local console, such as “Unfold the blue shirt.” Select Send or press Shift+Enter to start. The robot requests further actions from updated images. Select Robot Stop when you need to stop.
Keep the task objects in the camera's view. Start with a simple task and check the movement direction and gripper behavior on your setup.
Where to go
| Entry point | Purpose |
|---|---|
| Developer Console | Manage your account, API keys, model access, and usage |
| Model service Base URL | Configure your model service endpoint and call the cloud API |
http://127.0.0.1:8080/ on the control computer | Connect devices, view cameras, run and stop your robot |
The developer console manages your cloud account. Robot operation runs on your own control computer. To automate it, use the local HTTP API or MCP tools.
To connect another robot or develop a complete hardware adapter, continue with Connect new hardware and Hardware adapter interfaces.