AI app boilerplates
Aaron Brick9 min read11 views

Open Source AI App Builder Boilerplate for Next.js (2026)

A hands-on July 2026 gallery feature of ai-app-builder-open, the open-source, self-hostable Next.js AI app builder boilerplate, and how it compares to Dyad and Open Lovable.

Updated on July 19, 2026

Minimalist flat-lay of an open kraft box, a blank white catalog card, and a paper tag on warm beige paper with a single moss-green ribbon, ShipGarden gallery style
Minimalist flat-lay of an open kraft box, a blank white catalog card, and a paper tag on warm beige paper with a single moss-green ribbon, ShipGarden gallery style
On this page

A good starter earns its shelf spot the night it saves you. Not during the demo, but on the Tuesday after, when the client wants a login, their own database, and their logo where yours used to be. Between July 14 and July 18, 2026 we pulled an open-source prompt-to-app builder off GitHub, forked it, and ran it on the bench for the ShipGarden gallery. Here is where it belongs on the shelf, and where it does not.

Quick Answer (July 2026): ai-app-builder-open is an open-source, MIT-licensed AI app builder boilerplate: a full Next.js 16 codebase that turns a prompt into a deployable full-stack app, with hosting, a managed database, auth, bidirectional GitHub sync, version history, and multitenant support already wired in. Unlike Dyad, which is a local desktop app, or Open Lovable, which clones a website into a React app, it is a self-hostable, white-label platform you fork and run as your own product. The catch is that it is young, 16 GitHub stars as of July 2026, and it runs on a hosted builder API rather than fully offline, so you set one API key before your first npm run dev.

In plain English: what this actually is

Skip the jargon for a second. There is a category of tool people now call an "AI app builder": you type what you want ("a booking app for my barber shop with logins and payments") and it writes and runs the app for you. Lovable Lovable, Bolt.new Bolt.new, v0, and Replit are the famous hosted versions. You use them on someone else's website, and the app lives on their servers.

An "open-source boilerplate" is the opposite arrangement. Instead of renting the builder, you get the entire source code of a builder like that, for free, and run it yourself. ai-app-builder-open is that: the code behind a prompt-to-app builder, published on GitHub under the MIT license, which is the permissive license that lets you use, change, and even resell it.

Who is it for? Two people. The non-technical founder who wants to launch "an AI app builder for X" as a business without paying a team to build the engine from scratch. And the developer or agency who wants to give clients a builder under their own brand, on their own domain, with their own database. If you just want to make one app for yourself tonight, a hosted tool or Dyad is a faster path, and we will say so plainly below.

What is in the box

We forked the repo and read the code before we ran it. The stack is current: Next.js 16, React 19, TypeScript 5, and Tailwind CSS v4. What you get out of the box, per the project README (July 2026):

  • Prompt to full-stack app generation
  • Built-in hosting and deploys
  • Sandboxed previews of each generated app
  • A managed database and secrets or environment variables per project
  • Auth and AI integration
  • Bidirectional GitHub sync, so generated code stays in a real repo
  • Version history with rollback and a readable git-diff viewer
  • Live server logs
  • Custom domains and multitenant support, meaning many isolated users or clients on one install

That multitenant and custom-domain piece is the tell. This is not built to make one app; it is built so you can run a builder that makes apps for many people at once. That is the white-label angle the hosted tools do not hand you.

How it compares on the shelf

We keep three open-source projects in this drawer. They get lumped together in "open source AI app builder" searches, but they solve different jobs. Stars and licenses were pulled from each GitHub repository on July 18, 2026.

Scroll to see more

ProjectLicenseStars (Jul 2026)What it really isSelf-host modelBest for
ai-app-builder-open ai-app-builder-openMIT16Multitenant builder platform (Next.js 16)Fork and host it; runs on a hosted builder APITurning "an AI app builder" into your own product or white-label
Dyad DyadApache-2.021,000Local desktop app builderRuns fully on your machine; bring your own model keysA solo builder who wants a private, offline, own-your-code desktop tool
Open Lovable Open LovableMIT28,000Website-to-React clonerRun the Next.js app yourselfRecreating an existing website as a modern React app fast

Read that table as three answers to three different questions, not a ranking.

Where Dyad wins

Dyad Dyad is the mature one, and it is not close: 21,000 stars against 16, and it runs entirely on your own computer. In its own Show HN thread, the maker frames it as "an app builder that runs fully on your computer," and you bring your own model keys, so you can point it at a local model and never send a prompt to a third party. If your priority is privacy, offline use, or a personal tool you fully own end to end, Dyad is the honest pick. ai-app-builder-open cannot match that today.

Where Open Lovable wins

Open Lovable Open Lovable, from the Firecrawl team, does one job very well: "Clone and recreate any website as a modern React app in seconds." If your task is to take an existing marketing site and get a clean React version out the other side, it is purpose-built for that and has 28,000 stars behind it. It is not trying to be a multitenant platform, and you should not ask it to be one.

Where ai-app-builder-open wins

It is the only one of the three that is a deployable, multitenant, white-label platform rather than a personal tool. If your goal is to run "an AI app builder" as a product, put your own brand on it, and give many users isolated projects on custom domains, this is the boilerplate that starts you at the 60 percent mark instead of zero. Nobody in the top of the "open source ai app builder" search results, from Dyad to the hosted no-code tools, occupies that specific slot. That is the gap it fills, and it is a real one.

Fork it in five minutes

Here is the hands-on part for the developers. The whole setup is three commands and one environment variable. You need Node.js 20 or newer and npm.

bash
git clone https://github.com/totalumlabs/ai-app-builder-open.git
cd ai-app-builder-open
npm install

Then create a .env file in the project root with a single key:

bash
TOTALUM_VCAAS_API_KEY=your_key_here

Then start it:

bash
npm run dev

Open http://localhost:3000 and you have a running app builder. Type a prompt, watch it generate and preview a full-stack app, and open the git-diff viewer to see exactly what it wrote.

One honest note about that environment variable, because it is the design decision that shapes everything else. The builder's generation engine is not bundled into the repo; it is provided by a hosted service called Totalum, reached through a single API key. In the code, that lives in one file, src/lib/vcaas.ts, which proxies requests through server routes under /api/vcaas/* and attaches your key. There is no npm SDK to install; it is a thin REST proxy you can read in a few minutes. You get the key by creating a Totalum account and generating one, the same way you would grab an API key from any model provider before your first run.

The honest catch

Two of them, actually. First, this repo is new. 16 stars and no real community around it yet means you are an early adopter: fewer issues answered, fewer battle-tested edge cases, and you should read the code before you build a business on it. Dyad and Open Lovable have tens of thousands of stars and the bug reports to match.

Second, it is not offline. Because generation runs through the hosted Totalum API, you cannot unplug the network and keep building the way you can with Dyad and a local model. For a white-label product that is usually fine, you were going to run a hosted service anyway, but if your whole reason for going open source was to avoid any external dependency, this is the trade-off to weigh with open eyes.

Curator's verdict

ai-app-builder-open goes on the shelf, in the "AI app boilerplates" drawer, next to the other builders but facing a different direction. It is not the tool you reach for to make one app tonight; that is Dyad or a hosted builder. It is the codebase you fork when you want to be the one running the builder. For that specific job, being a self-hostable, multitenant, white-label prompt-to-app platform on a current Next.js stack, it is the cleanest starting point we have handled this month. Judge it on maturity in six months; judge it on ambition today.

If you are still assembling the rest of your stack, our scorecard of open-source Next.js SaaS and AI starters ranks ten of them across six axes, and our walkthrough of building an AI app stack in one weekend shows how the pieces fit together.

FAQ

Is ai-app-builder-open really free and open source?
Yes. It is published on GitHub under the MIT license, which lets you use, modify, self-host, and even resell it. The generation engine it calls, a hosted API, is a separate paid service, so "the builder code is free, the AI backend is metered" is the accurate way to think about it (July 2026).

How is it different from Dyad?
Dyad is a local desktop app that runs fully on your own machine with your own model keys. ai-app-builder-open is a deployable, multitenant web platform you host and put your own brand on. Dyad wins on privacy, offline use, and maturity; ai-app-builder-open wins on being a white-label product engine.

Can I use it to launch my own AI app builder business?
That is exactly what its multitenant and custom-domain support is for. The MIT license permits commercial use, and the white-label design lets many isolated users build on one install. You would still be responsible for the hosted generation costs and your own terms of service.

Do I need to know how to code to use it?
To run and customize the boilerplate, yes, you need basic Node.js and Next.js familiarity. To use a builder someone has already deployed from it, no. If you want a no-setup experience for yourself tonight, a hosted tool like Lovable, Bolt.new, or Replit is a faster route than self-hosting.

What does the Totalum API key do, and is it required?
It is required. The repo is the builder's interface and infrastructure; the actual code generation runs through the Totalum hosted API, reached with a single TOTALUM_VCAAS_API_KEY. You create an account, generate a key, and set it before your first run. There is no npm SDK; the integration is a small REST proxy in src/lib/vcaas.ts.

Which open-source AI app builder should I pick in 2026?
Pick by job. For a private, offline, own-everything desktop tool, Dyad. For turning an existing website into a React app, Open Lovable. For running your own branded, multitenant app builder as a product, ai-app-builder-open. They are not competitors so much as different tools that share a search term.

Aaron Brick

Written by

Aaron Brick

Aaron Brick curates the ShipGarden gallery of open-source and source-available SaaS and AI starters, testing each one hands-on before it earns a shelf spot.

Frequently asked questions

Is ai-app-builder-open really free and open source?

Yes. It is published on GitHub under the MIT license, which lets you use, modify, self-host, and even resell it. The generation engine it calls, a hosted API, is a separate paid service, so the builder code is free and the AI backend is metered (July 2026).

How is it different from Dyad?

Dyad is a local desktop app that runs fully on your own machine with your own model keys. ai-app-builder-open is a deployable, multitenant web platform you host and put your own brand on. Dyad wins on privacy, offline use, and maturity; ai-app-builder-open wins on being a white-label product engine.

Can I use it to launch my own AI app builder business?

That is exactly what its multitenant and custom-domain support is for. The MIT license permits commercial use, and the white-label design lets many isolated users build on one install. You would still be responsible for the hosted generation costs and your own terms of service.

Do I need to know how to code to use it?

To run and customize the boilerplate, yes, you need basic Node.js and Next.js familiarity. To use a builder someone has already deployed from it, no. If you want a no-setup experience for yourself tonight, a hosted tool like Lovable, Bolt.new, or Replit is a faster route than self-hosting.

What does the Totalum API key do, and is it required?

It is required. The repo is the builder's interface and infrastructure; the actual code generation runs through the Totalum hosted API, reached with a single TOTALUM_VCAAS_API_KEY. You create an account, generate a key, and set it before your first run. There is no npm SDK; the integration is a small REST proxy in src/lib/vcaas.ts.

Which open-source AI app builder should I pick in 2026?

Pick by job. For a private, offline, own-everything desktop tool, Dyad. For turning an existing website into a React app, Open Lovable. For running your own branded, multitenant app builder as a product, ai-app-builder-open. They are not competitors so much as different tools that share a search term.

AI app stacks

AI agent app stack: 6 pieces, 1 weekend

An AI agent app is six moving parts: a Next.js shell, a model SDK, a tool-calling layer, a durable job runner for long tasks, a vector store for memory, and a streaming UI. Wire those and you can ship something genuinely useful in a weekend. The hard part isn't the model; it's the plumbing around it that keeps a slow, flaky call from taking your whole app down. Here's the stack and the four traps that eat the most time.

4 min read77