AI-Assisted Connector Development

Build OpenMetadata connectors faster using AI skills with Claude Code, Cursor, Codex, and other AI tools

AI-Assisted Connector Development

OpenMetadata provides an AI skills toolkit that helps you build connectors faster using your preferred AI coding tool. The skills package includes:

• Scaffold tool — Generate JSON Schema, Python boilerplate, and an AI implementation brief with a single command
• Golden standards — 17 connector development standards that AI agents load as context
• Review skill — Multi-agent code review against OpenMetadata patterns
• Research agent — Automated API/SDK discovery for the source you’re building

The AI-assisted approach complements the manual development guide. The scaffold tool generates the same files you would create manually, and the standards encode the same patterns described in the manual guide.

How It Works

OpenMetadata uses a schema-first architecture. You define one JSON Schema for your connector’s configuration, and that single definition cascades through 6 layers:

JSON Schema (you define this)
    ├── Python Pydantic models     (make generate)
    ├── Java models                (mvn install)
    ├── TypeScript types           (yarn parse-schema)
    ├── UI config forms            (RJSF auto-renders from schema)
    ├── API request validation     (server uses Java models)
    └── Test fixtures              (tests import Pydantic models)

The scaffold tool generates the JSON Schema and all Python boilerplate. For SQLAlchemy database connectors, the generated code is nearly complete. For all other connector types, the tool generates skeleton files and a CONNECTOR_CONTEXT.md file that tells the AI agent exactly what to implement.

Supported AI Tools

Tool	Features
Claude Code	Full skill support: slash commands, parallel sub-agents, SessionStart hooks
Cursor	Skills as project rules, standards loading
GitHub Copilot	Context files, instruction-based
OpenAI Codex	AGENTS.md-based, standards as context
Windsurf	Rules-based, standards loading
Any AI agent	Read CONNECTOR_CONTEXT.md + reference connector

Quick Start

1. Install the skills

   Follow the installation guide for your AI tool.

2. Scaffold the connector

   source env/bin/activate
   metadata scaffold-connector
   

   This generates all boilerplate files and a CONNECTOR_CONTEXT.md implementation brief. See the scaffold guide for details.

3. Build with your AI tool

   Point your AI agent at the generated context file:

   Read ingestion/src/metadata/ingestion/source/{type}/{name}/CONNECTOR_CONTEXT.md
   and implement all TODO items. Use the reference connector as a pattern.
   

   See the full build workflow for each tool.

What’s in the Skills Package

skills/
├── .claude-plugin/plugin.json       # Plugin manifest
├── commands/                        # Slash commands (/scaffold-connector, /connector-review, /load-standards)
├── agents/                          # Sub-agents (researcher, validator, comment-checker)
├── standards/                       # Golden standards
│   ├── main.md                      # Architecture overview
│   ├── patterns.md                  # Error handling, pagination, auth
│   ├── testing.md                   # pytest patterns
│   ├── code_style.md               # Python and JSON Schema conventions
│   ├── schema.md                   # Connection schema patterns
│   ├── connection.md               # BaseConnection vs function patterns
│   ├── service_spec.md             # ServiceSpec registration
│   ├── registration.md             # How to register a connector
│   ├── performance.md              # Pagination, batching, rate limiting
│   ├── memory.md                   # Memory management, OOM prevention
│   ├── lineage.md                  # Lineage extraction methods
│   ├── sql.md                      # SQLAlchemy patterns, URL building
│   └── source_types/               # Per-service-type patterns
│       ├── database.md, sql_databases.md, data_warehouses.md, nosql_databases.md
│       ├── dashboard.md
│       ├── pipeline.md
│       ├── messaging.md
│       ├── mlmodel.md
│       ├── storage.md
│       ├── search.md
│       └── api.md
├── connector-building/SKILL.md      # Scaffold skill definition
├── connector-review/SKILL.md        # Review skill definition
└── load-standards/SKILL.md          # Standards loading skill