Generation levels

Sketch: direct generation

How it works. Sketch is the shortest path from an idea to a working application. You describe what you want, in a sentence or a paragraph: that's all. Maker interprets your request and calibrates the application to what it warrants: neither artificially inflated nor stripped down. No questionnaire, no steps: you write, Sketch builds.

Ideal for. Testing an idea before going deeper, getting a simple tool without devoting a meeting to it, showing rather than explaining. For those who want to see their need take shape immediately, and then decide whether it deserves more.

Use cases. An activity log, a small inventory, a bookings register, a personal dashboard. The tools you'd never have commissioned, but use every day once they exist.

Under the hood. A simple entry point changes nothing about the rigour of the output. Like every level of Blueprint Maker, Sketch rests on the same principle: the AI designs, it doesn't improvise your code. The interpreted structure of your request is turned into an application by the same deterministic builders as Craft and Masterpiece, database, screens, navigation, sample data. A simple application, but built like a large one.

AI on board. Sketch draws on Haiku, the fastest model in the Claude range, the one whose speed of interpretation matches the level's promise: a light request, an immediate answer, without mobilising more intelligence than the need calls for.

Availability & pricing. Sketch is available from the free Découverte plan, which includes 5 credits. Each generation consumes 1 credit, shown before you launch. It's the entry point to Blueprint Maker: enough to judge on the real thing, with no commitment.

Craft: a few scoping exchanges

How it works. Craft adds what a plain description lacks: your answers. Before generating, Maker asks a few targeted questions, a few minutes, no more: who the actors in your activity are, which indicators you track, which sequences pace your work. Each answer sharpens the specification. The application it produces no longer describes a domain in general, it describes yours.

Ideal for. Tools that must speak your language: entities named as in your trade, indicators you actually track, workflows that mirror your sequences. For those who know that between a generic CRM and their CRM, the difference comes down to ten good questions.

Use cases. A sales tracker that talks about mandates and viewings rather than “deals”, an association's membership management with its dues and reminders, a stock tracker with your own alert thresholds. The same kind of application as elsewhere, but tuned to your terms, your measures, your habits.

Under the hood. Craft enriches every stage of the pipeline without giving up its principle: the AI designs, it doesn't improvise your code. Your answers calibrate a richer specification, entities confirmed with you, indicators defined from your answers, whose business consistency is checked before generation. The same deterministic builders then produce the application, down to the sample data, made relevant to your domain rather than generic. More relevance in, the same reliability out.

AI on board. Craft engages Sonnet, the balanced model in the Claude range, to lead the exchange and establish the specification, where understanding your trade is decided. Haiku supports it on ancillary tasks, including producing believable sample data. Already the logic of the whole range: each model in its place.

Availability & pricing. Each Craft generation consumes 3 credits, shown before you launch. With the Pro plan (€25/month, 60 monthly credits), that's up to 20 personalised applications a month, the core use of Blueprint Maker.

Masterpiece: deep scoping

How it works. Masterpiece begins where others stop: with a scoping led as a functional consultant would. The exchange unfolds in four stages: your data model (entities, relations, attributes), your journeys and screens, your indicators with their calculation formulas, and finally your business rules and access rights by profile. At each stage, a validation point: you review, amend, approve before moving on. The application isn't generated from an interpreted sentence, it's built from a complete specification you have approved. It's the work of a professional project scoping, compressed into a conversation.

Ideal for. Masterpiece is for applications meant for production, not prototypes: multi-role with differentiated rights, precise management rules, indicators that are computed and not merely displayed, edge cases documented rather than ignored. For those who run a daily working tool, who know the devil is in the rules, and that the rules must be written before they're coded.

Use cases. A CRM aligned with your real sales cycle rather than a generic pipeline, tracking a farm with its seasons and plots, a multi-site commercial management, a lightweight workshop ERP with its production orders. Wherever the gap between “an app for the domain” and “your app” is felt every day, Masterpiece is the level that closes it.

Under the hood. Like every level of Blueprint Maker, Masterpiece rests on one principle: the AI designs, it doesn't improvise your code. The validated specification is turned into an application by deterministic builders, the same specification produces the same application, structured, testable, reproducible. Where your trade departs from the standard, Masterpiece additionally engages a chain of specialised agents: the first identifies the bespoke screens your activity warrants, the second puts them to the test, usefulness, consistency, redundancy, the third generates them, the fourth validates the code produced. Creativity is summoned only where it's needed, and never without control.

AI on board. Masterpiece mobilises two Claude models, each in its place: Sonnet establishes the complete specification and structures the application; Opus, built for the most demanding reasoning, is engaged exclusively on the bespoke screens, where your trade resembles no other. The right intelligence in the right place, on a foundation that doesn't vary.

Availability & pricing. Masterpiece is reserved for the Max plan (€149/month, 400 monthly credits). Each generation consumes 10 credits, up to 40 bespoke applications a month, and the cost is shown before every launch. No surprises: you know what you're committing before you generate.

At a glance

The three levels compared.

How to choose

Start with the most direct level that meets your need, then adjust: you can always regenerate at a higher level. Each level's credit weight is published, you know the cost before you launch.