Catalog¶

Reusable platform assets—volumes, tables, models, and operators—are registered in a single object tree called the Catalog. It is not a list of data: knowledge bases and agents use it to locate data, and it also locates reusable operators and operators published as APIs.

Object hierarchy¶

The Catalog organizes assets in an object tree:

Workspace → Catalog → Database → Volume / Table

Catalogs are either MOI Catalogs, where data is stored in the platform, or external database Catalogs, which connect to an external database while keeping data at the source.
A database is a container for related data objects. It can contain volumes and tables.
Tables store structured data, while volumes manage unstructured files such as documents and images. Models and custom operators are other managed Catalog assets.
Tables and volumes can coexist in one database, so structured and unstructured data can be managed together. Models and custom operators are also managed in the Catalog.

Where data is stored and what the Catalog records¶

After data enters MOI, it has two physical destinations: original files are stored as bytes in object storage, while structured outputs from workflows are written to MatrixOne. Table data lives in MatrixOne and volume files live in object storage; the two data types are physically separate.

The Catalog records metadata and relationships for these objects, including asset type, source, processing workflow, and AI-use relationships. Knowledge bases and agents use the Catalog to obtain information about the data assets they reference.

Data lineage¶

The Catalog records data lineage automatically. Each platform processing step explicitly declares its input and output: data import declares “external source → table,” workflow operators declare “table → processing → table,” and knowledge bases declare which tables they contain. These declarations form a lineage graph that lets any processed output be traced back to its original source.

Two types of Catalogs: managed and federated¶

	MOI Catalog	External database Catalog
Data location	In MOI (object storage and MatrixOne)	In the source database; MOI records metadata only
Write access	Data can be created, updated, and deleted	Read-only; the upstream system controls its structure
Queries	Run within the platform	SQL is pushed down to the source database; data does not enter MOI

An external database Catalog brings existing enterprise databases, such as MySQL, PostgreSQL, and Oracle, into a unified Catalog view and query experience without migration.

Usage boundary¶

Catalog registration reflects operations performed through the platform. If you change MatrixOne database tables directly outside the platform, synchronize metadata before relying on the Catalog to match the actual data.