Skip to main content

AI Toolkit

The AI Toolkit is an integration layer over the normalized REST API. It helps developers connect golf data to models without designing a separate grounding format. For clients that support the Model Context Protocol, the API also exposes these capabilities through the native MCP endpoint.

Tools​

GET /v1/ai/tools returns model-compatible function definitions for supported search and retrieval operations.

Use tools when the agent should decide when and how to query the API.

search_golf_clubs returns each matching club's id (its stable UUID), slug and name. For a detail lookup, pass that UUID to get_golf_club or get_golf_club_context as identifier:

search_golf_clubs → result.id → get_golf_club({ identifier: result.id })

The detail-tool schema accepts either identifier or the legacy slug property. Prefer identifier with the UUID returned by search. The slug property remains supported for backwards compatibility; do not use it for new agent workflows.

For ordinary place requests, the search tool uses near, an optional ISO country code and radius_km. The API resolves the place and performs the geographic search. town only matches the postal town stored on a club record.

Compact context​

GET /v1/ai/context/clubs/:identifier returns a token-efficient club representation. Use the UUID returned as id from search as identifier; an existing unique slug is also accepted for backwards compatibility. It follows the same Fairway or Tour data-access policy as the regular club endpoint.

Use compact context when the model needs to summarize, compare or recommend a known club.

Prompt definitions​

GET /v1/ai/prompts lists stable task IDs, versions, variables and output schemas. GET /v1/ai/prompts/:id returns one definition.

Prompt mechanics remain language-neutral. Localized wording is supplied through the translation catalogues.

Initial tasks include:

  • Club summary
  • Visitor briefing
  • Club comparison
  • Club recommendation
  • Natural-language search

Structured output​

Output schemas make recommendations and comparisons easier to validate. They also force agents to report unknown information instead of inventing an answer.

Model independence​

The Toolkit does not require one AI provider. Tool definitions, context and schemas can be adapted to models with tool-calling support. MCP-compatible clients can instead discover supported tools and prompts directly through the MCP endpoint.