DEV Community

Germey
Germey

Posted on

Build an AI API Site You Can Monetize: Customer Key Onboarding

An AI API business is not only a collection of model endpoints. Customers also need a place to discover services, understand pricing, read documentation, create credentials, and verify that a request was billed correctly.

Ace Data Cloud's White Label API Platform packages those customer-facing pieces into a branded developer portal. You can configure a site name, logo, service catalog, pricing presentation, documentation, support information, and an optional custom domain. It is intended for developers who will integrate through an API key; the separate White Label Studio is aimed at people using AI tools through a web interface.

This guide focuses on one operationally important workflow: onboarding a customer from registration to a verified first API call. It does not create a site, change DNS, open an account for a customer, or make a payment.

Disclosure: This article is based on Ace Data Cloud's public documentation and was prepared with AI assistance for an Ace Data Cloud content program. It is an independent technical rewrite, not a claim that I operated a production tenant or earned revenue from one.

1. Make the catalog support one clear first use case

Before inviting customers, choose a small number of services that you understand and can support. In Site Management → Manage → Services and Pricing, configure the site-wide markup and display currency, then add a per-service override only when necessary.

A per-service markup replaces the site-wide value for that service; it is not added on top. Leaving the override blank inherits the site-wide setting, while 0 explicitly means no markup. The display currency is a presentation setting, so the service detail page and final checkout remain the places to verify the actual transaction.

Services and Pricing settings for markup, currency, and service coverage

Treat catalog visibility as merchandising, not access control. Hiding a service does not automatically hide its documentation, and neither action necessarily prevents an existing valid key from calling an endpoint. Configure the service catalog and documentation catalog separately; enforce credential scope through the credential controls that are actually available for the customer's application.

For a beginner-friendly documentation path, put these items first:

  1. What the service does.
  2. How the customer obtains a key.
  3. One minimal request.
  4. Pricing and required parameters.
  5. Errors, limits, and asynchronous result polling when applicable.

Do not rename public model identifiers, package names, environment variables, or protocol fields into marketing terms. A friendlier display title is useful, but executable values must continue to match the service documentation.

2. Give the customer four distinct values

Most onboarding mistakes come from mixing up the website and the API. Provide a short handoff containing exactly these fields:

Portal URL: https://portal.example.com
API Base URL: https://api.example.com
Service / model: <public identifier from the service documentation>
Documentation: https://portal.example.com/docs/<service>
Enter fullscreen mode Exit fullscreen mode

The portal URL is where the customer registers, logs in, reads docs, and manages credentials. The API Base URL is where the application sends requests. If custom domains are enabled, obtain the required CNAME targets from your own console; never copy a target from another tenant or tutorial screenshot.

The example values above are placeholders. They are intentionally not real credentials or tenant addresses.

3. Keep identity and credentials customer-owned

Ask the customer to enter through your portal URL and register or log in using one of the authentication methods enabled for the site. After authentication, the browser should return to the branded portal.

A white-label portal should not be described as a separately delivered gateway or an independent tenant identity system. The documented implementation uses a shared identity system. Existing users can log in, while a first-registration attribution test requires a genuinely new account.

Next, the customer opens the portal console's application area, selects the required application, and enters credential management. The documented routes are:

/console/applications
/console/credentials
Enter fullscreen mode Exit fullscreen mode

The customer creates and stores their own API key. The site operator's key must never be shared across customers. Where the form supports them, use a recognizable key name plus an expiration time, usage limit, and allowed-API scope appropriate to the test.

Credential creation form with name, usage limit, expiration, and allowed APIs

For application configuration, keep secrets server-side:

AI_API_BASE_URL="https://api.example.com"
AI_API_KEY="<customer-owned-key>"
AI_SERVICE="<documented-public-service-or-model>"
Enter fullscreen mode Exit fullscreen mode

Do not commit .env files, put a key in browser JavaScript, paste it into support chat, or include it in screenshots. Separate keys by customer, project, and environment where practical. If support is needed, send the timestamp, request path, HTTP status, and a safe request identifier—not the full authorization header.

4. Verify one small request end to end

Copy the smallest example from the selected service's documentation. Change only the fields the documentation requires: the Base URL, the customer's key, and necessary request parameters. Check the current price before submitting, especially for media jobs.

Then verify both sides of the operation:

  • The request returns a successful result, or an asynchronous task reaches its terminal success state.
  • The customer's usage history contains a matching record with the expected service, time, and consumption.

For image, music, or video workflows, an accepted task ID is not a completed generation. Follow the service-specific polling instructions until the task succeeds or fails. “One successful operation plus one matching usage record” is a stronger acceptance criterion than merely receiving an initial HTTP response.

If the key was created only for acceptance testing, the customer can disable or delete it after confirming that no application still depends on it. A low-risk follow-up request should then be rejected, proving that revocation took effect.

5. Connect onboarding to monetization—accurately

A portal can support a commercial AI service, but registration and API usage alone do not guarantee operator income. Customer attribution and revenue are separate concerns.

The public documentation distinguishes default attribution from forced attribution. Forced attribution can override referral links, invitation cookies, and defaults, so it should be enabled only when that business rule is intentional. Test first-registration attribution with a new account and record the entry domain, registration time, and resulting relationship without exposing personal data.

Revenue depends on qualifying completed purchases and the applicable distribution rules. Refunds, risk controls, order status, settlement timing, and upstream sharing can affect the final record. A markup changes the channel price, but the difference between displayed platform price and customer price should not be presented as guaranteed operator earnings. Likewise, there is no basis for promising fixed daily income or a commission on every API call.

A sound commercial acceptance test therefore compares three records: the completed customer order, the customer's balance and usage, and the operator's distribution record. That links acquisition, valid purchase, and service delivery without turning a technical guide into an earnings claim.

Launch checklist

Before sharing the portal publicly, verify:

  • the public catalog, price, and checkout agree;
  • documentation contains authentication, errors, limits, and async polling where needed;
  • portal and API domains are not confused;
  • every customer creates and protects their own key;
  • hidden catalog items are not assumed to be access controls;
  • a minimal call produces a matching usage record;
  • attribution and completed-order records are checked separately from API calls;
  • support, terms, and privacy links are visible.

This workflow is deliberately narrow, but it is enough to turn a branded API site from a collection of pages into an onboarding path a developer can actually complete.

Sources

Top comments (0)