Build with AI

The Databricks Partner AI Dev-Kit is an open-source toolkit that gives your AI coding assistant the patterns and rules needed to generate PWAF-compliant Databricks connectors. Instead of reading the docs and writing code manually, you describe what you need and your AI assistant generates a connector with authentication, telemetry, and compliance built in.

The kit works with Claude Code, Cursor, Codex, and any MCP-compatible AI coding assistant.

What you can build

Use case	Connector	Languages
Run SQL queries against a SQL warehouse	SQL drivers and connectors	Python, Java, Go, Node.js
Call workspace APIs (jobs, Unity Catalog, clusters)	Databricks SDKs	Python, Java, Go
Run Spark workloads from outside Databricks	Databricks Connect	Python
Make direct HTTP calls in any language	REST API	Any
Use an ORM against a SQL warehouse	SQLAlchemy	Python

Every generated connector covers all four authentication flows and includes built-in User-Agent telemetry.

How it works

Step 1 — Install Clone the repository and run the installer, or connect via MCP server (no clone required).

Step 2 — Pick a skill Each connector type has a dedicated skill. Tell your AI assistant which skill to use, or let the MCP server serve it on demand.

Step 3 — Generate Run the connector's build prompt. Your AI assistant reads the skill, generates the connector code, configures authentication, sets the User-Agent string, and writes the test runner.

Step 4 — Validate Run the automated PWAF compliance check. A fully compliant connector scores 12/12.

Getting started

Option A — MCP server (no cloning required)

Install the dependency:

pip install mcp

Add to your .mcp.json (Claude Code) or ~/.cursor/mcp.json (Cursor):

{
  "mcpServers": {
    "databricks-pwaf": {
      "command": "python",
      "args": ["-m", "databricks_pwaf_mcp"],
      "cwd": "/path/to/databricks-partner-ai-dev-kit/mcp"
    }
  }
}

Skills are fetched from GitHub automatically on first use and cached for 24 hours.

Option B — Clone and reference directly

git clone https://github.com/databricks-solutions/partner-ai-dev-kit.git databricks-partner-ai-dev-kit
cd databricks-partner-ai-dev-kit
bash install.sh

Point your AI assistant at the skills/ directory. For Claude Code, add to CLAUDE.md; for Cursor, add to .cursor/rules/.

Supported stacks

Language	Connectors
Python	SDK, SQL Connector, SQLAlchemy, Databricks Connect
Java	SDK, JDBC
Go	SDK, SQL Driver
Node.js	SQL Driver
Any	REST API

Authentication coverage

All skills generate connectors that implement all four authentication types required for PWAF validation:

Auth type	Use case
Personal Access Token (PAT)	Simple, user-specific access
OAuth M2M	Machine-to-machine, service principals, backend systems
OAuth U2M	Interactive, browser-based, user-delegated access
Token pass-through	Pre-obtained tokens, CI/CD, headless environments

The auth type is selected at runtime via APP_AUTH_TYPE — never DATABRICKS_AUTH_TYPE.

PWAF compliance check

Every generated connector ships with an automated compliance check that validates all PWAF requirements:

validate_pwaf_tool("/path/to/your/connector")

A fully compliant connector returns:

{
  "overall": "PASS",
  "summary": { "PASS": 12, "FAIL": 0, "WARN": 0, "SKIP": 0 }
}

The checks cover User-Agent telemetry, all four auth types, isolated test runner, no hardcoded credentials, .env.template present, and a build report.

What's next

• Review integration requirements: Understand what Databricks requires for partner validation. See Integration Requirements.
• Explore telemetry attribution: Learn the User-Agent format the Dev-Kit implements for you. See Telemetry & Attribution.
• Get the toolkit: Clone, install, and run your first build prompt. See Partner AI Dev-Kit.