** AI 驱动的 Databricks 平台工程。** 配置工作区、配置 Unity 目录、搭建网络、管理组和权限——所有这些都通过自然语言与你的编码代理（Claude Code、Codex、Cursor、Copilot、Gemini 等）进行。

Databricks AI 开发套件的兄弟：该套件为编码代理提供 Databricks 专属技能，用于在平台上构建，而该套件则专注于搭建平台本身——工作区、Unity 目录、身份以及跨 Azure、AWS 和 GCP 的网络。

观看概览

为什么大家都在使用Databricks AI平台套件？ ——简要介绍该套件的功能及使用方法。

它的作用

告诉你的经纪人你需要什么，剩下的由他们帮你处理：

• Workspace 配置 — VNet/VPC 注入，跨 Azure、AWS、GCP 的安全集群连接
• Unity 目录 — 主观设置：自管理元存储、每个环境目录、外部位置、奖章模式、湖屋联合
• 身份与治理 — 账户级 SCIM 组、分级 RBAC、作业服务主体
• 工作空间配置 — SQL 仓库、集群策略、秘密范围、IP 访问列表

Databricks Platform Kit

AI-powered Databricks platform engineering. Provision workspaces, configure Unity Catalog, set up networking, manage groups and permissions — all through natural language with your coding agent (Claude Code, Codex, Cursor, Copilot, Gemini, and more).

Sibling to the Databricks AI Dev Kit: where that kit gives coding agents Databricks-specific skills for building on the platform, this kit focuses on standing up the platform itself — workspaces, Unity Catalog, identity, and networking across Azure, AWS, and GCP.

Watch the overview

Why Is Everyone Using the Databricks AI Platform Kit? — a short walkthrough of what the kit does and how to use it.

What it does

Tell your agent what you need, and it handles the rest:

• Workspace provisioning — VNet/VPC injection, Secure Cluster Connectivity across Azure, AWS, GCP
• Unity Catalog — opinionated setup: self-managed metastore, per-env catalogs, external locations, medallion schemas, Lakehouse Federation
• Identity & governance — account-level SCIM groups, tiered RBAC, service principals for jobs
• Workspace config — SQL warehouses, cluster policies, secret scopes, IP access lists
• Private networking — private link (AWS), Private Endpoints (Azure), Private Service Connect (GCP), hub-spoke, NCC for serverless connectivity (when you need it)
• Built-in verification — a 3-path test (classic cluster + serverless SQL warehouse + serverless notebook job) designed to catch common configuration failures after a deploy (it does not guarantee correctness or fitness for any particular use)

The agent writes the Terraform and SDK calls from scratch based on your specific requirements — no rigid templates to fill in.

Architecture

.claude/skills/                      # 6 focused skills
  platform-provisioning/             #   Workspace creation + deploy
    SKILL.md                         #     Shared workflow + interaction guidance
    AZURE.md / AWS.md / GCP.md       #     Cloud-entry navigation files
    {azure,aws,gcp}-1-auth.md        #     Auth setup per cloud
    {azure,aws,gcp}-2-deploy.md      #     Deploy patterns per cloud
    {azure,aws,gcp}-3-gotchas.md     #     Sharp edges per cloud
  unity-catalog-setup/               #   Metastore, catalogs, governance
    SKILL.md + AZURE.md / AWS.md / GCP.md
  identity-governance/               #   Groups, users, SPs, RBAC
    SKILL.md                         #     Cloud-agnostic
  workspace-config/                  #   SQL warehouses, policies, secrets
    SKILL.md                         #     Cloud-agnostic
  private-networking/                #   Private link, hub-spoke, NCC, PSC
    SKILL.md + AZURE.md / AWS.md / GCP.md