Guide / How it works

The screen is yours.
Start with one useful thing.

CAR-TER turns a layout into native readings and controls on your iPhone or iPad. You can try a ready-made template, connect a project, or ask an assistant to help build the interface.

Jump to building with an assistant

How a layout works

1 of 3

A layout puts useful controls together.

A reading, a chart, a switch: each is a control. Arrange them into a layout, and use tabs when you need another view.

Practice scene only. No device is connected.

This desk layout has a temperature reading and a light switch.

Read the whole walkthrough

  1. A layout

    A reading, a chart, a switch: each is a control. Arrange them into a layout, and use tabs when you need another view.

  2. A source

    A reading can come from a public API, a phone sensor, or a service you connect. A template can already have this connection set up.

  3. An action

    The service handles the request and reports what happened. A tap is a request—not proof that a device changed. Read-only layouts simply let you explore their data.

Same app, different starting points

Try first.
Connect when you need to.

A template can be ready to use.

The public-data collection already has its feeds configured. Open one, inspect the readings, and check the source time. You don’t need to run a hub to get started.

Choose a public-data template →

Your project may need a connection.

For a custom remote, a service handles the real work: reading a sensor, controlling a light, or querying your software. CAR-TER is its interface—not a promise that every device connects automatically.

Make it together

Describe. Preview.
Make one change.

An agent is an assistant that can use tools, not just write a reply. With CAR-TER’s authoring tools and your permission, it can draft a layout, show it on a paired device, and check what rendered.

Build with an assistant

1 of 5

Say what you want to see and change.

“Make a desk dashboard with room temperature and a light switch. Keep it readable on my phone, and label any sample values.” The assistant should ask where the real information and commands will come from.

Illustrated workflow—not a running agent. Nothing here pairs, connects, or saves.

Your idea → a small, specific brief.

Developer view of this step

Start with the intended controls, data source, units, action behavior, and phone size. Never invent a working device connection.

Read the whole walkthrough

  1. Describe

    “Make a desk dashboard with room temperature and a light switch. Keep it readable on my phone, and label any sample values.” The assistant should ask where the real information and commands will come from.

  2. Draft

    It consults CAR-TER’s control documentation and drafts the arrangement. You can also build the same layout with Python and carterkit; using an assistant is optional.

  3. Preview

    In a Studio Session, the assistant can push the draft to your paired app. Check the size, labels, and arrangement. A preview made with sample data is still a preview.

  4. Check

    The authoring tools can read back the layout and control values from the device. Compare them with the intended design, then test the actual data and action connection separately.

  5. Refine & keep

    “Make the temperature larger and move the switch below it.” The assistant revises the draft. Save when you are happy, keeping your original if you want a comparison.

Set up the authoring toolsPrefer Python? Start with carterkit →

The illustrated path uses an external AI client with carter-mcp. Client/provider setup and charges are separate from CAR-TER. You can use ready-made layouts without an AI service.

Names for things you’ve now seen

A small map of the parts.

You don’t need to memorize this. Come back when a word is useful.

Control
One piece of the interface: a chart, a reading, a button, or a switch.
Layout
The arrangement of controls and tabs, plus how they receive information and send actions.
Template
A ready-made layout you can try or use as the start of your own version.
Source
Where information comes from: an API, a phone sensor, or a connected service. A fresh retrieval can still contain an old observation.
Hub
Your project’s service or program. It supplies readings and handles commands. Direct API and sensor layouts don’t always need one.
Relay
The message carrier for a connected session. It delivers messages between participants; it isn’t your light-control logic or your AI assistant.
Studio Session
A paired authoring session in which tools can send a draft to the app and read back what rendered.
carter-mcp & carterkit
The assistant’s authoring tools and the Python building library. Both help create the layout the app runs.

Technical architecture · Data sources · Control reference

One useful screen is enough.

Try a template first. The rest can wait until you have something you want to change.

Open your first layout

Website motion uses Anime.js, hosted here under its MIT license. All practice scenes use sample values.