Skip to main content
The backend is a Next.js 15 application that handles API requests, stores data in Convex, and runs background jobs through Inngest. This guide walks you through getting all three services running locally and building for production.

​
Prerequisites

  • Node.js 20+ and npm
  • Accounts and API keys for the required services below
The following environment variables are required. The backend will not function without them.Stripe and RevenueCat keys are optional — only needed if you want payment processing.

​
Local development setup

1

Clone the repository

Clone the repo with all submodules. The mobile app and Expo template are git submodules, so --recurse-submodules is required.
2

Install dependencies

3

Configure environment variables

Copy the example file and fill in your API keys:
Open .env.local and fill in at minimum the six required variables listed above. See the environment variables reference for the full list.
4

Deploy the Convex schema

This pushes your database schema to your Convex deployment. Run this once on first setup, and again any time you change convex/schema.ts.
5

Start the Inngest dev server

Inngest handles background jobs — spawning sandboxes and running the AI agent. Start it in a separate terminal:
The Inngest dashboard is available at http://localhost:8288. Make sure INNGEST_DEV=true is set in your .env.local.
6

Start the Next.js dev server

In another terminal, start the Next.js server:
The backend runs at http://localhost:3000.Alternatively, use npm run start to start both Next.js and Inngest together with a single command.

​
Production build

To build a production-optimized bundle:
Then start the production server:
In production, set INNGEST_DEV=false (or remove the variable) and configure Inngest to use a real event key from your Inngest dashboard.

​
Updating the Convex schema

Whenever you change convex/schema.ts or any file in the convex/ directory, re-deploy your Convex functions:

​
Next steps

Once the backend is running, build the E2B sandbox template so the AI agent has an environment to run in: