How the Platform Works
Every application built on Anybuild follows the same four-phase lifecycle. AI is involved in most of these phases, but it is never the only option - anything AI can do, a Creator can also do directly with point-and-click tools or code.
The lifecycle: Specify → Model → Generate → Evolve
Section titled “The lifecycle: Specify → Model → Generate → Evolve”Specify - describe what the application should do. A short prompt, a detailed written specification, or an uploaded requirements document (DOCX, up to 50MB) all work as a starting point. The platform’s AI turns this into a structured specification, organised entity by entity, and asks clarifying questions about anything it’s uncertain of.
Model - review and refine the data model the AI generates from the specification, shown as an interactive entity relationship diagram. This is the structural blueprint of the application: every entity, field, and relationship. You can edit entities directly on the diagram or keep using AI chat to make broader structural changes - both work side by side, and there’s no requirement to use AI here if you already know exactly what you want.
Generate - once the model is approved, the platform builds a complete, working application from it: database, forms, list views, navigation, a branded login screen, default security, and the Draft/Submit record workflow. This happens automatically - there is no separate build or deployment step.
Evolve - the generated application is a starting point, not a finished product. From here you keep building in the Designer: add business logic with workflows and server functions, configure security roles, build custom views and UI, connect external data sources. This is where most of the ongoing work happens, and it’s also where external AI tools come in - see below.
Three ways AI shows up in Anybuild
Section titled “Three ways AI shows up in Anybuild”Internal AI does the heavy lifting during Specify and Model, and stays available throughout Evolve. Beyond generating the initial specification and data model, the same AI can be used to build SQL views, UI components, and other artefacts on request - you describe what you want and refine it through chat rather than building it from scratch by hand.
AI in applications is AI embedded inside the running application itself, not just in the design-time tools. Creators can wire AI into their data model and business logic - using it for things like document processing (OCR), image recognition, classification, and AI-assisted routing decisions inside workflows. In effect, Creators get an AI interface they can use as a building block when defining application logic, the same way they’d use a SQL expression or a scalar function.
External AI comes in during the Evolve phase, once an application already exists. Creators can use tools like Cursor, Claude Code, and Codex to extend the application via the platform’s API. Anybuild validates, normalises, and absorbs what comes back from these tools, so changes still have to respect the model’s structure and rules - external AI accelerates the work, but the platform is what governs it.
In short: AI generates, the platform governs. Whichever mode you’re using, the data model is the source of truth that everything else has to be consistent with.
No-code, with a code option
Section titled “No-code, with a code option”Anybuild is built around defining entities and their relationships at a high level of abstraction, instead of hand-coding the lower-level pieces - statements, variables, tables, screens, servers, security - that a traditional application needs. This is the core of why building on Anybuild is dramatically faster than coding the same system from scratch: a Creator working at the entity level is doing in minutes what would otherwise take hours or days of low-level implementation.
That said, code is never off the table. Anybuild supports defining components either through point-and-click designers or directly with code, and the two are fully interoperable - something defined with code becomes a reusable building block in the point-and-click designers, and vice versa.
- SQL is available for views, functions, and field expressions - it’s a highly productive language for data queries, analysis, and manipulation, and the platform leans on it directly rather than abstracting it away.
- HTML, CSS, and JavaScript are available for custom UI components and pages, for Creators who want full control over presentation. This is entirely optional - the platform ships with a large library of ready-made components and layouts that cover most needs without writing any code.
The result is a platform that defaults to no-code for speed, but never forces a rebuild in a different tool when something genuinely needs custom logic or a custom interface.