Shanc Agent Search
Search your data context in plain language
Agent Search runs over your context layer. People and agents ask in natural language — about a metric, an entity, a rule — and get the governed definition: how it’s calculated, what it depends on, where it comes from. One answer, the same for everyone who asks.
- Natural-language search across metrics, entities and business rules
- Every result carries its calculation logic, dependencies and source
- Same search, same answers, for people and for agents
How it works
One place to ask
The definition, without the archaeology
Today the answer to “which revenue is board revenue” lives in someone’s head, a dbt model, and three Slack threads. Agent Search puts it behind a single question — the layer already holds the resolved definition, and search is how you reach it.
Answers with provenance
Every result shows its working
A search result is a definition, not a guess: the exact calculation, the tables and entities it depends on, and the source it was built from. Trust comes from being able to check — every answer is inspectable down to the logic.
Built for agents too
Precise retrieval, small payloads
Agents use the same search over MCP: a question about churn retrieves the churn definition and its dependencies — nothing more. Precise retrieval keeps context windows small and answers grounded, which is where the accuracy comes from.
Finding the right definition is the slowest step in every data question
53% of data experts receive three to four follow-up questions after fulfilling each ad-hoc request — the queue regenerates faster than it clears. Search that returns the governed definition answers the follow-ups before they’re asked. (Sigma Computing survey)
Technical integration challenges are declining across data teams (35% → 27%), while data ownership ambiguity and quality gaps persist nearly unchanged. The bottleneck has moved from connecting data to agreeing what it means. (dbt Labs, State of Analytics Engineering 2026)
42% of data practitioners name data not being where business users need it as their top concern. Definitions locked in models, catalogs and tribal knowledge are the sharpest version of that gap. (dbt Labs, State of Analytics Engineering, 2023)
When the definition is one search away, the data team stops being the search engine.