Salesforce Launches Catalog To Ground Agents In Org Metadata
Salesforce announced Salesforce Catalog on November 10, 2026, a Data 360 powered metadata index that gives AI assistants and agents structured context about how an org is built. Catalog covers Salesforce objects, fields, flows, Apex, permission sets, users, and Data 360 assets including data streams, calculated insights, and segments. The feature grounds answers, code generation, and metadata changes in the org’s current configuration rather than generic platform knowledge.
Catalog grounds agents in org-specific metadata
Catalog indexes metadata and relationships as the org changes. An assistant can search by meaning as well as exact names, so a request such as “where do we track whether an order arrived?” can resolve to an org-specific field such as `Shipment.Delivery_Status__c`. The assistant receives the closest matches and relevant details instead of scanning the entire org or relying on an outdated metadata export.
Access follows the signed-in user’s permissions. Each Catalog question runs as that user, allowing an assistant to expose object and process relationships before an admin removes a field or a developer generates code against the wrong API name.
Enablement requires Data 360
Salesforce Catalog is available in all editions supported by Data 360. An admin enables it in Data Cloud Setup under Feature Manager, and the required permission is Data Cloud Architect. The enablement documentation says the first activation starts a full metadata sync, followed by incremental sync for supported metadata types and a daily full refresh.
Production usage is billed as Data 360 usage. Salesforce’s Catalog billing documentation says pre-production orgs such as Developer Edition, sandboxes, and preview orgs are not billed for Catalog access through the Metadata Grounding MCP server, while production orgs are billed for Data 360 usage. Teams connecting custom AI clients can use the Metadata Grounding MCP Server with OAuth. The service is currently in beta and provides separate production and sandbox endpoints.
