Beauty API for Agencies: Delivering Beauty Personalization for Clients Without Building a Model

Agencies working with beauty clients get the same brief in different wrappers: "we want something personalized for the launch", "can the site tell people their undertone", "we need a try-on for this campaign". The deadline is a media calendar, not a research roadmap. A beauty API for agencies exists for exactly that gap — it lets a digital, CRM or performance agency deliver skin analysis, shade logic and makeup previews as an integration, instead of pitching a model-training project nobody in the room has budget for.

This article is about how to scope, build and hand over that kind of work without over-promising.

What agencies are actually being asked for

Strip the brief down and there are usually three requests hiding inside.

The first is diagnosis: the shopper wants to know something about themselves — skin tone and undertone, skin type and condition, hair attributes — before they trust a recommendation. The second is preview: the shopper wants to see a product on their own face before they add it to cart. The third is guidance: the shopper wants a shortlist from this brand's catalog, not a generic article.

These are three different builds. Treating them as one "AI experience" is how timelines slip. Separate them in the proposal and each one gets a clear owner, a clear integration point and a clear way to be measured.

What a beauty API for agencies actually has to deliver

An agency has constraints a brand's in-house team does not: multiple clients, short engagements, and code that someone else may inherit.

Self-serve access, so discovery is not a procurement project

The fastest way to de-risk a proposal is to call the API before you write it. With MaIA Scan, a developer creates an account, buys credits and gets an API key — then sends a selfie and receives a beauty profile as JSON: skin tone and undertone, skin type and condition, hair attributes, with a confidence score per field. You can validate the response shape against the client's actual creative before the SOW is signed.

That matters commercially. You stop pitching an abstraction and start pitching a response you have already seen, on photos that look like the client's real audience. The model is trained on Brazilian skin, which is the difference between a demo that holds up in a São Paulo focus group and one that quietly fails on half of it.

Per-client separation

Agencies run several brands at once, sometimes competing ones. Palettes for makeup try-on are calibrated to each brand's real products inside a project, so one client's shade library never leaks into another's. Keep API keys per client, not per agency, and make the key part of the handover document from day one — it avoids the awkward conversation when the retainer ends.

Credits you can forecast

Usage-based access is friendlier to campaign work than a platform contract: a six-week activation consumes what it consumes. Estimate volume from the media plan — expected sessions times the share you expect to engage with the tool — and price the build around that. Do not quote a conversion number to the client. Any uplift figure you have read belongs to the brand that published it, and results vary with each catalog, each traffic source and each channel.

Mapping the three building blocks to deliverables

Diagnosis: the beauty profile

The profile is the data layer everything else sits on. For an agency, the common deliverables are a landing page quiz that ends in a real analysis instead of self-declared answers, a CRM enrichment flow that writes attributes back into the client's base, or a product-page module that pre-filters shades. The JSON is plain enough to drop into a tag manager payload, a serverless function or the client's existing personalization engine.

One rule to write into the deck: this describes visible attributes of a photo. It is not a diagnosis, it does not treat anything, and copy that implies otherwise should be rewritten before legal sees it.

Preview: makeup try-on

MaIA Try-On comes in two forms, and choosing the right one is an agency decision. The API is photo-based: you send a photo, the lipstick or foundation is rendered on it, and the image comes back — ideal for campaign pages, WhatsApp flows, social landing pages and anywhere you want a shareable result image.

The browser SDK, in early access since September 2026, covers thirteen products — lipstick, foundation, blush, face powder, under-eye concealer, eyeshadow, mascara, eyebrow makeup, eyeliner, bronzer, contour, highlighter, lip liner — plus a combined Full Look, in live camera mode or photo mode, with any #RRGGBB colour, self-hosted by the brand. Live camera mode renders locally in the browser, so nothing is uploaded. Access is granted per account, so confirm availability before you put the full product list in a pitch; the API alone renders lipstick and foundation on a photo.

Guidance: the advisor

When the brief is conversational — WhatsApp, chat, a guided experience on product pages — MaIA Beauty Advisor reads the beauty profile and recommends the brand's own products. It is delivered as a project with the B4A team rather than a self-serve API, so scope it as a joint workstream with catalog preparation on the client's side, not as a two-sprint front-end task.

Scoping the engagement

A workable shape for a first project: week one, key, test calls and a decision on which products are in scope. Week two, a working prototype on the client's real catalog and real photos, including low-confidence handling — what the interface says when lighting is poor and the result is uncertain. Week three onward, production integration, analytics events and the handover pack: endpoints used, key ownership, credit balance, and who to contact when the client wants a second brand added.

The agencies that keep this work are the ones that hand over something the client's own team can operate. Document it like you are leaving.

Ready to scope a client build? Create an account and get an API key in the API documentation, or talk to the B4A team about a multi-brand setup.