Google adds Looker metric rules to the data catalog for agents

Google Cloud released LookML indexing in its Knowledge Catalog for Looker Core. LookML is the Looker language that defines how a business metric, such as net revenue, is calculated. The catalog holds these formulas, and AI agents read them and do not guess. Google tells you the sync is on by default. It gave no test results.

How it works

The catalog tells an agent which data is in it and how it is governed. LookML adds the correct formulas, the connection paths and the limit rules. Google tells you that updates to production LookML show in the catalog in a short time. Gemini Enterprise reads them with read only skills. Developers who make custom agents use the LookupContext API.

The proof

Google gave no test results or precision numbers. It names two errors that agents make on raw tables. One is conflicting formulas for one metric. The other is connection errors that multiply rows and inflate totals.

What is not known

Google does not tell how much more correct agents become. The release is for Looker Core on Google Cloud. Google tells you that no single tool can hold the business context and the metric math.

Insight: Vendors race to own the context layer for agents

Google indexes metric formulas in its catalog by default. On 20 October 2026, Neo4j released a Context Engine, a map of company data for agents. Each vendor looks like it wants its own layer to be the one that agents read first. The one that holds the metric definitions can set the net revenue number that agents use.