Home

Lantern documentation

Using Lantern requires two installed pieces: a package in a Next.js app and a companion running on each user's Mac. Lantern's relay connects them. The user approves the exact site before it sends supported AI requests to Apple Intelligence, Codex, or Claude on that Mac.

Apple Intelligence performs inference on the Mac. Codex and Claude use their signed-in online services. In both cases, the web request still passes through the app and Lantern's relay before reaching the Mac.

Start here

Add Lantern to a Next.js app

Build a Connect Lantern flow, then send AI SDK requests to a model on each user’s Mac.

Use Lantern on your Mac

Check the requirements, install Lantern, approve a site, and understand what that approval allows.

SDKs

Vercel AI SDK

Send AI SDK requests through each user’s Lantern.

React SDK

Add the Connect Lantern button and remember the user’s choice.

Images

Generate or edit images through Codex on the user’s Mac.

Request path

See how your server reaches the Mac stored in this browser’s connection.

Core concepts

How Lantern works

Follow one request from your app to the user’s Mac, then back again.

Supported models

How Lantern uses Apple Intelligence, Codex App Server, codex exec, and Claude.

Reference

Local APIs

Use Lantern’s HTTP and WebSocket interfaces from tools on the same Mac.

Configuration

See Lantern, approved-app, provider, and session settings.