Skip to content

MCP Tools

MCP Tools are callable functions that let agents interact with your semantic layer. Unlike Resources, which are read-only, tools accept parameters and return computed results.

Search across models, fields, glossary terms, and business rules by keyword or natural-language query.

ParameterTypeRequiredDescription
querystringYesThe search term or natural-language question

Example use: An agent receives the question “What is our monthly churn rate?” and calls context_search with query: "churn rate" to find relevant models and metrics.


Returns a plain-language explanation of a model, field, or glossary term, including its business context, relationships, and usage notes.

ParameterTypeRequiredDescription
namestringYesThe identifier of the model, field, or term to explain

Example use: An agent calls context_explain with name: "orders.net_revenue" to understand what net_revenue means, how it is calculated, and how it relates to other fields.


Validates a proposed SQL query or expression against the semantic layer, checking for correct joins, appropriate grain, and adherence to business rules.

ParameterTypeRequiredDescription
sqlstringYesThe SQL query or expression to validate

Example use: Before returning a generated query to the user, the agent calls context_validate to confirm the query respects the defined grain and does not violate any guardrails.


Returns the current tier status and scorecard, showing which rules pass and which require attention.

ParameterTypeRequiredDescription
tierstringNoThe tier level to check (bronze, silver, gold). Defaults to the highest achieved tier.

Example use: An agent calls context_tier with tier: "gold" to see which Gold-level requirements are not yet met.


Returns the golden queries defined for a model. Golden queries are curated, known-correct SQL examples that demonstrate proper usage of the model.

ParameterTypeRequiredDescription
modelstringYesThe model identifier

Example use: An agent calls context_golden_queries with model: "orders" to retrieve reference SQL it can adapt for the user’s question.


Returns the guardrails defined for a model. Guardrails are SQL-based constraints that generated queries should respect — for example, always filtering by a specific date range or never aggregating non-additive measures incorrectly.

ParameterTypeRequiredDescription
modelstringYesThe model identifier

Example use: An agent calls context_guardrails with model: "orders" to learn that queries against the orders model must always include a WHERE order_date >= ... filter.