Glossary term
Application metadata
Application metadata is the structured, typed definition of a business application — its objects, fields, relationships, views, permissions, workflows, actions, and APIs — stored as source files that a runtime reads and executes directly, rather than as generated application code.
Also known as app metadatametadata-driven application definition
In practice
The distinction that matters is not "config versus code" but what happens at run time. Generated code is a one-way export: once the generator has produced a controller, a form, and a migration, those artifacts drift apart and the original intent is gone. Metadata stays the live source. Change the permission rule in the definition file and the API, the UI, the audit log, and the agent tool all change with it, because all four are derived from that one declaration on every request.
That property is what makes metadata the right target format for AI. A complete CRM expressed as metadata fits in roughly 150k tokens, so a coding agent can hold the whole system in one context window instead of guessing at a codebase it can only sample. And because the definitions are typed and schema-validated, a wrong field name or an unknown permission value fails at the validation gate rather than shipping as a runtime surprise.
It is also what makes AI-written software reviewable. When an agent adds an approval step, the diff is a handful of readable lines in a definition file — not four hundred lines across a controller, a template, a migration, and a test. A reviewer with business authority but no interest in framework internals can read that diff, understand what changed about who may do what, and sign off.
Where this term is used
Pages and articles that put this term to work.
Product pages
- Data modeling Objects, relationships, validations, and formulas become tables, migrations, a query language, and REST APIs — with an object designer, an ER diagram, and datasource sync that federates existing databases in place.
- Platform ObjectOS is the commercial production platform for ObjectStack apps, combining in-app AI, deployment, and team operations with the open, governed ObjectStack runtime.
- Agent developer Teach a coding agent the open ObjectStack target format, Strict TypeScript definitions, Zod schemas, validation gate, and whole-context proof needed to generate reviewable applications.
Articles
- D’une phrase aux objets, vues, permissions et Agents : la chaîne de génération d’AI Builder L’important dans AI Builder n’est pas de transformer une phrase en pages, mais de décomposer une demande métier en objets, champs, vues, workflows, permissions, automatisations et outils d’Agent.
- Combien de tokens fait une application métier ? Un CRM complet sous 150k Un CRM complet tient sous 150k tokens de métadonnées typées : logique métier sous 100k, UI autour de 50k. Le CRM de référence fourni reste à environ 16k.
- Pourquoi le low-code atteint ses limites Le low-code accélère pages et workflows, mais les systèmes métier complexes dépendent d'objets, permissions, intégrations, changement et maintenabilité.