Skip to content

MCP Resources

MCP Resources are read-only URIs that return structured context about your semantic layer. Agents can read these to load model definitions, glossary terms, and tier scorecards without calling any tools.

Returns the full built manifest — every model, glossary term, business rule, golden query, and guardrail in your context directory. This is the complete picture of your semantic layer in a single read.

When to use: When an agent needs a comprehensive understanding of the entire data domain before starting work.

Resource URI: context://manifest

Returns the full definition of a single model, including its fields, grain, relationships, business rules, golden queries, and guardrails.

When to use: When an agent is working with a specific model and needs its complete definition without loading the entire manifest.

Resource URI: context://model/orders
Resource URI: context://model/customers

The {name} parameter corresponds to the id field in your model YAML files.

Returns all glossary terms and their definitions, including synonyms, related terms, and canonical usage notes.

When to use: When an agent needs to understand business terminology — for example, to disambiguate “revenue” vs. “net revenue” or to know that “ARR” is a synonym for “annual recurring revenue.”

Resource URI: context://glossary

Returns the tier scorecard for a specific tier level, showing which rules are required, which pass, and the overall compliance status.

When to use: When an agent needs to understand the quality and completeness of the semantic layer at a given tier.

Resource URI: context://tier/bronze
Resource URI: context://tier/silver
Resource URI: context://tier/gold

Returns a blank AI Blueprint template — an OSI-compliant YAML starting point for a new data product.

When to use: When starting a new data product from scratch. The template includes all OSI sections with comments explaining each field.

Resource URI: context://data-product/template

Returns the AI Blueprint for a data product — the complete semantic specification as portable OSI YAML. Includes the model’s schema, governance metadata, glossary terms, metrics, relationships, business rules, guardrail filters, and golden queries — everything an AI agent needs to work with this data correctly.

When to use: When you need to share a data product with another team, feed it to an AI agent, migrate to a different warehouse, or create a portable backup of your semantic layer.

Resource URI: context://data-product/orders
Resource URI: context://data-product/customers

Agents typically start by reading context://manifest or a specific context://model/{name} resource, then use Tools like context_search or context_explain for interactive exploration. The resources provide the foundation; the tools provide the interaction.

For portability, agents can read any data product’s AI Blueprint via context://data-product/{name}, or scaffold a new one from the blank template at context://data-product/template.