PRODUCT ENGINEERING / CONNECTED BUSINESS SYSTEMS
Omnistu Global - Everything ConnectedStart a project

Engagement scope - proposed scope

Choosing a White Label AI App Builder Provider

This page describes proposed scope for a build in this category, agreed in writing per project. Nothing here is a finished product, a fixed bundle or a fixed price.

Scope areas for this kind of build

What You Arrange Before Day One

Open your LLM provider accounts

  • Generation runs on provider accounts you open and pay for, under each provider's own terms. Omnistu has no model of its own, so pick which of the 22 you will offer at launch and have the keys ready.

License the in-browser runtime

  • Projects run inside StackBlitz WebContainers. Using them in production for a for-profit product needs a commercial license from StackBlitz, which you obtain directly. StackBlitz sets that price, not us.

Pick a payment provider

  • The platform meters credits but takes no payments itself. Stripe or a similar gateway is wired in during deployment, so the account should exist before that step comes round.

Typical scoping questions

Who owns the source and the user data?

  • Ask for the repository, the schema and the user and project records by name. A right to use a workspace is not ownership, and the difference shows the first time you want to change a plan.

What happens when you stop paying?

  • On a reseller plan the workspace usually stops with the invoice. On a platform you own, you keep running it with your hosting and provider bills and nothing else. Get it in writing.

Who licenses the in-browser runtime?

  • If projects run in the browser through StackBlitz WebContainers, production use in a for-profit product needs a commercial license from StackBlitz. Ask who obtains it. Here, you do, and we tell you before you buy.

Whose LLM accounts pay for each prompt?

  • Generation should run on provider accounts you open, under their terms, with keys injected on the server by a proxy. A key that ever reaches the user's browser is a key you will have to rotate.

Where is the list of what is not included?

  • Ours names it: no mobile app, no payment processing, and CSP, HSTS, two-factor sign-in and encryption at rest as deployment work. A quote without exclusions was never costed against your plans.

How many providers, and how hard is the next one?

  • "Multi-model" sometimes means two. Ask for the list, then ask what adding one more involves. Here it is a class behind the shared provider contract and one registry entry.

The Six-Step Delivery Process

Your brand, then your servers

  • Product name, logo, colors, theme and domain applied across the workspace and desktop builds, then deployment to your cloud. Execution happens in the browser, so the server footprint stays modest.

Keys and the model catalog

  • Your provider keys added on the server, the model list populated, and input and output credit rates set per model on every plan against what your providers actually bill you.

Plans, credits and the gateway

  • Free, Pro and Enterprise priced with their allowances and model access, the free tier capped on purpose, and your payment provider connected with webhook signature checks.

Red Flags That Mean Walk Away

Use this list on us as well. With a metered AI product, most of the expensive surprises are commercial, not technical.

What a White Label AI App Builder Has to Get Right

Builds that run off your servers

  • Installs, builds and previews execute in the user's browser tab, so hosting costs do not climb with each active builder and user code stays outside your network.

Metering from real token counts

  • Each call is priced on its actual input and output, checked against the balance on the server and deducted there, where a tampered client cannot reach it.

A ledger that settles arguments

  • Grants, spends and promotional credits are appended with amount, balance, model and reason. On a metered product, that record is what turns a dispute into a short conversation.
  • More than 30 starters across eight categories load a complete working project. A blank workspace is where builders lose new users before they spend a credit.

An exit for the user

  • Deploys to GitHub, GitLab, Vercel, Netlify or Cloudflare Pages, or a ZIP download. People commit more work to a tool they are free to leave.

Custody, Licenses and Limits

A buyer's security review asks where the code goes and who can see it. A buyer's lawyer asks whose licenses cover what. Both answers belong in front of you before the purchase.

Self-Hosted, Browser-Executed

  • The platform runs on servers you control, and generated projects execute inside each user's browser rather than in a shared remote sandbox. No third-party SaaS that you do not choose sits between your users and their work.

Keys Held by the Operator

  • Provider keys are kept by you and added to requests on the server by the Express auth proxy, never returned to the browser. Rotating or revoking a key is one console action, shown as a masked preview.

The Licenses Are Yours

  • We supply software, not rights to a model or a runtime. LLM usage runs on accounts you open under each provider's terms, the commercial StackBlitz license for WebContainers is obtained by you, and payments run through your own gateway.

Stated Limits

  • Each capability carries one of three labels, Included, Supported or Configuration required, so what you hold after purchase matches what you were shown before it.

Illustrative internal AI-builder requirements

A proposed internal AI-assisted coding scenario, not an enterprise client engagement or a deployed platform. Requirements below need source, integration, security and workload verification.

Challenges

  • Define which code and prompts may reach each provider or service, with explicit source/data boundaries; no private-code containment is certified by this preview.
  • Evaluate per-team usage metering, budget limits, charge reconciliation and overrun behavior; predictable costs are not promised.

Goal

  • Assess login, metering and administrative controls required for the internal builder, distinguishing existing inspected source from development still needed.
  • Scope eligible model providers, permitted usage, price configuration and staff permissions; verify provider terms and current costs separately.
  • Require traceable usage-to-team/model records with reconciliation, retries and access-boundary tests.

Proposed implementation options to verify

  • Browser-side project execution is a recorded architecture option. Verify package/runtime compatibility, isolation, networking and any server dependencies before choosing it.
  • Scope plan/credit controls and protected provider credentials, with explicit permissions and actual security-test evidence.
  • Propose operator configuration for eligible models, prices and feature flags, subject to implementation and permission tests.
  • Express and a Cloudflare workerd runtime are recorded options, not an inspected deployment or final stack choice.

Bring the checklist to every bidder

Omnistu is an independent software development project. We are not affiliated with, connected to, sponsored by, or endorsed by Bolt.new , StackBlitz or any other AI app builder or code generation service.

White label AI app builder describes a category of product, not any one company. Brand names appear elsewhere on this site only to describe the kind of platform being built and the terms buyers search for.

Discuss this scope by message

Tell us your market and audience, and we will respond with a practical read on scope.

team@omnistu.com