AI that works for you at home.

Set up AI agents to handle the tasks you never get to. They work in the background — and ask before doing anything important.

Get Started

How It Works

1

Set up agents on your home computer

Install Hearth on any Mac or Linux machine at home. Configure AI agents for the tasks you want handled — adjusting the thermostat, controlling the lights, running the sprinklers, researching purchases, drafting emails, and more.

2

They work. You approve what matters.

Your agents get busy. When they're about to do something sensitive — access a file, make a network request, run a command — they ask you first. You get a notification on your iPhone and tap Allow or Deny. Takes three seconds.

3

Build trust. Reclaim your time.

Tap "Always Allow" on actions you're comfortable with and Hearth remembers. Over time your agents get more autonomous for the things you trust them with, and interruptions become rare. You get the results without the busywork.

Everything you need to put AI to work

Talk to Your Agent

Send instructions to any agent from your phone, mid-task. Text or voice — no laptop required.

Approve from Anywhere

Your agents ask permission in real time. You approve or deny from your iPhone, wherever you are. Push notifications alert you even when the app is closed. Pro

Smart Rules

Teach Hearth what to approve automatically. After a few sessions, routine actions happen without interrupting you — only genuinely new or sensitive requests come through.

See Everything

The Chat tab shows exactly what every agent is doing, step by step, in real time. Nothing happens behind your back.

Multiple Agents

Run different agents for different jobs — one for research, one for email, one for files. Each has its own permissions and activity history.

Pull the Plug

Stop any agent instantly with one tap. You're always in control, no matter what.

Voice Input

Dictate instructions hands-free. On-device transcription via WhisperKit — nothing leaves your phone. Pro iPhone 12+

Rich Previews

Before you approve, see exactly what the agent wants to do — file contents, web requests, commands — in a clear, readable format.