Skip to content
Advertisement
Google Cloud launches Knowledge Catalog with Looker, Snowflake and SAP semantic model support

Google Cloud launches Knowledge Catalog with Looker, Snowflake and SAP semantic model support

The rebranded Dataplex successor wants to be the context engine that keeps enterprise AI agents from making things up

Google Cloud has launched Knowledge Catalog with expanded semantic model support from Looker, Snowflake, and SAP. It is a direct bet that the next enterprise AI bottleneck is context, not raw model horsepower.

The pitch is simple. AI agents are only as smart as what they know about your business. Google wants to be the company that tells them.

What Google actually shipped

Knowledge Catalog officially debuted on April 22, 2026. It is a rebrand and upgrade of what Google previously called the Dataplex Universal Catalog.

Google describes the product as a Gemini-powered “universal context engine.” Think of it as a reference librarian sitting between a company’s data and the AI agents trying to use it.

The catalog pulls together technical and semantic metadata from Google Cloud services and from partner platforms. Technical metadata covers things like table structures and data types. Semantic metadata covers meaning: what a “customer” or “active account” actually refers to inside a specific company.

Knowledge Catalog uses AI-driven insights to serve up what Google calls governed context. The goal is fewer hallucinations and more accurate agents. Teams reach that context through search and through APIs.

Advertisement

Later updates in October 2026 added more machinery. On October 22, Google introduced self-learning context enrichment and a revamped Knowledge Studio, both in preview.

The self-learning feature is designed to spot gaps on its own by watching how agents and users interact. It then pulls verified definitions from sources such as documents and repositories.

On October 26, LookML semantic language indexing reached general availability for Looker Core. The feature includes live syncing of changes.

LookML is the modeling language Looker customers use to define business metrics. Indexing it brings deterministic business logic into Google’s context graph.

The Snowflake and SAP angle

The partner integrations may be the most strategically interesting piece. Knowledge Catalog now works with Snowflake and SAP for governed data sharing and unified semantic context across systems.

Crucially, Google says this happens without replicating data. The integrations support zero-copy access, meaning the data stays where it lives while the catalog reads its meaning.

The benchmark claims

Google is making bold performance claims, and they come with an important caveat: the numbers are from Google’s own internal benchmarks.

According to Google, agents using Knowledge Catalog achieve nearly four times greater accuracy. The company also cites up to 33% lower latency and a 75% decrease in token usage.

Each figure targets a different pain point. Accuracy is about trust. Latency is about user experience. Token usage is about the bill.

What this means for enterprises and the cloud race

For enterprise data teams, the immediate question is practical. Companies already running Looker, Snowflake, or SAP alongside Google Cloud get a more direct path to feeding those definitions into AI agents.

The LookML general availability is the most concrete win today. Organizations that already invested years in building Looker semantic models can now reuse that work for agent context, rather than rebuilding it from scratch.

The self-learning enrichment and Knowledge Studio remain in preview. That means they are not yet production-grade commitments, and teams should treat them as features to test rather than infrastructure to depend on.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.