ERD Studio

AI-native data warehouse modeling for dbt.

Design your silver and gold layers on a visual canvas. Tell Claude Code what to build.
It writes the dbt models, schema YAML, and tests — no need to touch code or the diagram manually.
Purpose-built for Claude Code + dbt. Other AI assistants supported.

AI builds your models  ·  Physical stage from manifest  ·  Real-time drift detection  ·  AI sync reconciliation  ·  40MB+ manifests

How It Works

1. Point your AI assistant at the bronze layer. It profiles the source tables against your requirements and drafts a logical silver ERD on the canvas

2. Prompt the AI to materialise the silver ERD into dbt models, schema YAML, and tests

3. Refine your requirements and have the AI compose a logical gold ERD — facts, dimensions, and business-aligned grain — on top of the silver foundation

4. Prompt the AI to generate the gold dbt code, with tests derived directly from the ERD's keys and relationships

ERD Studio ships with a baseline AI harness that teaches your assistant the domain format and sync workflow. Bring your own skills, prompts, and style guides on top so the generated dbt code reflects your team's naming conventions, modelling patterns, and review standards.

Two Stages, One Truth

The physical stage is computed at runtime — no files on disk. Relationships come from dbt relationships tests, cardinality from unique tests.

Toggle the Diff button to compare stages. Discrepancies are color-coded directly on the canvas: matched extra missing type-mismatch cardinality-mismatch. Missing models appear as translucent ghost nodes.

Reconciliation

When design and warehouse drift apart, ERD Studio generates a sync plan — a JSON file mapping every discrepancy to a concrete fix. Choose which side is the ground truth per item, then let your AI assistant execute the plan: adding columns to YAML, writing dbt tests, updating domain JSON, or removing stale relationships.

AI Coding Harness

One command installs a schema reference that teaches your AI the domain format, naming conventions, and editing rules:

Cmd+Shift+P → dbt: Install AI Coding Harness

Once installed, describe what you want in plain English — your AI creates models, generates dbt schema YAML with tests, executes sync plans, and validates naming conventions. No manual editing required.

Features

• Drag-to-relate — long-press a column, drag to create FK relationships
• ELK auto-layout with manual repositioning (Shift+L)
• Model templates — dimension, fact, bridge, SCD2, or blank
• Domain tagging — auto-tags dbt YAML with domain:{name} for scoped builds
• Medallion layers — bronze, silver, gold, platinum, or custom
• Central model store — one YAML per model, referenced across domains
• Full undo/redo via VS Code WorkspaceEdit
• Handles 40MB+ manifests with worker-thread parsing

Getting Started

Prerequisites: VS Code 1.85+ • dbt project with dbt_project.yml • manifest.json in target/

1. Open your dbt project in VS Code
2. Click the ERD Studio icon in the Activity Bar
3. Initialize — follow the prompt to create the erd-studio/ folder
4. Create a domain — Cmd+Shift+P → dbt: Create Semantic Domain
5. Design — add models, define columns, drag to relate
6. Install the AI harness — Cmd+Shift+P → dbt: Install AI Coding Harness

Settings

Contributors

Jason Kwe Core concept, UI/UX, testing & iteration

Ginny Sync reconciliation, auto-layout, depth partitioning, testing & iteration

_{MIT License • Made for the dbt community}