Skip to content
openlaunch
Esc
↑↓navigate↵open⌘Jpreview
On this page

Get started

Sign in to your openlaunch console, pair a device and choose the functions your agent can use. Developers can also run the bridge locally.

Connect your first device

  1. Open your console and continue with Google.
  2. Choose Pair Pi 4 or Pair Uno R4. Your console creates a single-use enrollment that lasts 10 minutes.
  3. Follow the Pi installer or Uno USB setup. The device appears in your console when it connects.
  4. Open Agent permission, choose ChatGPT or Codex, select its functions and set an expiry. Connect the agent.

You can run a function yourself from the device card, or ask a linked agent to use an approved function. Inspect the action receipt for the result. To add your own hardware behavior, see Build your own functions.

Run a local developer console

You need Node 24+ and Git. On macOS or Linux, run:

curl -fsSL https://www.openlaunch.dev/install.sh | bash

The installer downloads the source, builds the console, and starts the bridge. It does not flash a board or change cloud resources. You can read the hosted script first. The GitHub copy is the backup; both are published from the same source commit.

Open http://127.0.0.1:8788. On macOS, your owner session key is already on the clipboard: paste it into the session field and select Connect. On Linux, copy the token value from .cache/local/owner.json in the installed checkout. Keep that file private. Your session key is a password for your console; never give it to an agent.

The default installation is ~/.local/share/openlaunch. An existing checkout is reused without resetting your files. To stop the bridge, press Ctrl-C. To start it again:

cd ~/.local/share/openlaunch
npm run start:local

Add your board

In the console, select Uno R4 WiFi or Pi and create an enrollment. Follow the Uno guide or Pi guide to connect it within 10 minutes.

For a real board, the bridge needs a trusted HTTPS address the board can reach. Your Mac’s loopback address only works on that Mac. See network setup.

Connect your agent

Follow agent connections. Then grant that agent health for your device in the console. Ask it to check the board’s health and inspect the result. Add LED or matrix access when you want the Uno to use those capabilities.

A grant is a permission slip: it names the agent, device, allowed capability, and expiry. Pairing alone doesn’t give an agent permission.

Work on the source

Already cloned the repository? Run npm run setup from its root. The installer needs only Node and Git. Building or testing the Pi agent also needs Go; flashing an Uno needs Arduino CLI.

Use npm run check for shared API, permissions and protocol tests. npm run test:e2e runs a simulated device through the local bridge; it requires Go and does not operate physical hardware. Platform details are in the Mac handoff.

Was this page helpful?