typeorm-docs-mcp

Offline TypeORM schema documentation tools for both humans and AI clients.

typeorm-docs-mcp analyzes TypeORM entity source files without connecting to a database. It produces a canonical schema graph, Mermaid ERD, Markdown documentation, and a missing-documentation audit. The same core implementation powers both the CLI and MCP tools.

Why this exists

This project is not meant to wrap work that an AI could do by reading files directly. The goal is to provide deterministic schema extraction and scriptable outputs that are useful on their own, while also giving AI clients stable schema context for database design review.

Related MCP server: Code Graph Context

Clean-room note

The project is inspired by ORM-to-Markdown and ERD documentation workflows, but it does not copy implementation code, file structure, tests, or README text from prisma-markdown.

CLI usage

typeorm-docs-mcp analyze --entities "src/**/*.entity.ts"
typeorm-docs-mcp erd --entities "src/**/*.entity.ts"
typeorm-docs-mcp markdown --entities "src/**/*.entity.ts" --title "Schema" --output ERD.md
typeorm-docs-mcp audit --entities "src/**/*.entity.ts"
typeorm-docs-mcp serve

Example ERD output

erDiagram
users {
  uuid id PK
  varchar email UK
}
blog_posts {
  uuid id PK
  uuid author_id FK
}
blog_posts }o--|| users : author

Example audit output

{
  "missingTableDescriptions": [],
  "missingColumnDescriptions": ["blog_posts.body"],
  "missingRelationDescriptions": []
}

MCP tools

Start the MCP server over stdio:

typeorm-docs-mcp serve

Available tools:

• analyze_typeorm_schema

• generate_mermaid_erd

• generate_typeorm_markdown

• find_undocumented_schema

Common input:

{
  "entities": "src/**/*.entity.ts",
  "projectRoot": "/absolute/path/to/project"
}

MVP limitations

• Offline source analysis only.

• No database connection or credential handling.

• No DataSource.initialize() live metadata mode yet.

• No migration diff analysis yet.

• No HTML documentation output yet.

Development

npm install
npm test
npm run build