Databricks Docs Server

Semantic search over Databricks documentation, API reference, Terraform provider docs, and knowledge base — as an MCP server for Claude Code.

Uses all-mpnet-base-v2 embeddings + ChromaDB with keyword boosting for high-relevance results.

Setup

cd tools/databricks-docs-mcp
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Build the Index

python indexer.py

Clones doc repos, scrapes docs.databricks.com + kb.databricks.com, chunks, embeds, and stores in chroma_data/. Takes ~15 min on first run (uses MPS GPU on Apple Silicon). Skips already-indexed collections on re-run.

Connect to Claude Code

Add to .mcp.json:

{
  "mcpServers": {
    "databricks-docs": {
      "command": "/path/to/tools/databricks-docs-mcp/.venv/bin/python",
      "args": ["/path/to/tools/databricks-docs-mcp/server.py"]
    }
  }
}

Run /mcp in Claude Code to connect.

Tools

Tool	What it searches
search_databricks_docs	General docs — Genie, Unity Catalog, warehouses, etc.
search_databricks_api	REST API & Python SDK reference
search_terraform_databricks	Terraform provider resources
search_databricks_genai	GenAI Cookbook — RAG, evaluation, chunking
search_databricks_cli	CLI commands, bundles, workspace ops
search_databricks_kb	Troubleshooting & known issues
search_all	All collections at once
research	Deep multi-query research on a topic
list_collections	Show collections and chunk counts
list_pages	List indexed pages in a collection
fetch_page	Get all chunks for a specific page

Doc Sources

Source	Collection	~Chunks
docs.databricks.com	databricks_docs	14,200
kb.databricks.com	databricks_kb	1,100
terraform-provider-databricks	terraform_databricks	2,200
databricks-sdk-py	databricks_api	107
genai-cookbook	databricks_genai	176
databricks CLI	databricks_cli	689