AI Catalog

One model for the whole estate.

A list of servers is an inventory. Governance needs something stronger: a model where roles, agents, tools, tables and columns are all described in the same terms, so a policy written once means the same thing everywhere.

The model

Thirty tables, one shape.

Everything AI Connect discovers lands in the same governance model. Because each connector resolves into that shared shape, adding a new source of AI activity does not change how policy is written.

Servers and what they expose

MCP servers, their tools, resources and prompts, with the side effect of each tool recorded.

Data down to the column

Tables, columns and datasets behind those servers, with sensitivity labels applied at discovery.

Who and what is calling

AI actors, the applications they belong to, and the models behind them.

Identity

Roles and role mappings synced from the identity provider your organization already uses.

Relationships

Access paths and lineage edges — not just which entities exist, but how they connect.

History

Versioned snapshots of servers, models and agents, so change is visible rather than overwritten.

What you browse

What the catalog holds.

The catalog is where you go to answer "what do we actually have" — before any question about who may reach it.

AI Actors

Every agent seen calling, with its review state, what it has reached, and when it was last active.

Tools

Every operation exposed across your servers, with its side effect and the server it belongs to.

Resources

The named resources your servers expose, and which servers expose them.

Prompt Templates

Prompts published by your servers — governable objects in their own right, not just text.

Data Assets

Tables, columns and datasets, carrying the sensitivity labels applied at discovery.

Versions

Snapshots of servers, models and agents, so the catalog shows a point in time rather than only the present.

Why a canonical model

The catalog is what makes policy portable.

A rule about a sensitive column means the same thing whichever server fronts that data, because both resolve into the same model. Governance does not have to be rewritten each time the estate grows.

The catalog holds the relationships between all of this. Seeing them — and governing the paths they form — is AI Control.

Talk to us

See the model against your own estate.

The catalog is most convincing with real entities in it. We are glad to walk through what that looks like for the systems you actually run.