AI Search MCP server

AI Search MCP server is a Databricks ai / ml capability, introduced June 2025.

Databricks managed MCP server that lets agents run semantic search over AI Search indexes to find relevant documents, with results governed by Unity Catalog permissions.

• The rename left fossils in the API - AI Search was formerly Vector Search, and the old /api/2.0/mcp/vector-search/ URL prefix and vector-search OAuth scope still work. 🔗
• The model asks the question but does not choose the retrieval strategy: preset _meta parameters pin the result count, filters, score threshold, reranked columns, and whether the search is pure vector or hybrid. 🔗

⚠ Limitations: Querying an index through the server requires Databricks managed embeddings. 🔗

Category

AI / ML

Introduced

June 2025🔗

Also known as

Vector Search MCP server

Verified

2026-08-20