The open-data protocol

fd-open-data-protocol is the manifest contract a datasource must expose to be ingested by fd-open-data-mcp. It’s what makes “one MCP, every source” possible: any provider that conforms becomes a first-class source in the ontology.

What a datasource manifest declares

A conforming datasource publishes a catalog describing:

• name — the provider’s stable identifier
• functions — the query operations it supports (e.g. fetch daily bars, fetch GDP)
• columns — the physical output fields of those functions
• entity definitions — the entity types it covers (stocks, countries, industries, …)
• relationships — links between entities
• concepts — the semantic indicators it can provide
• bindings — the mapping from its columns to ontology concepts

The MCP registers these into the ontology database; propose-bindings then proposes the column→concept mappings that the semantic layer uses to resolve your natural-language queries.

Why it matters

Because the contract is explicit, the MCP can rank candidate sources for a concept by quality × accessibility × freshness, fetch from the best one, fail over, and cache — all without per-source special-casing.

See the fd-open-data-protocol repository for the full manifest schema and validation rules.