Agent-ready metadata
Capture what each object is for, not just its technical shape. Rich extended properties and exports give downstream agents the semantic map they need to resolve ambiguous questions to the right tables, joins, and columns.
Ground your AI agent in real database meaning
Document SQL Server so your AI agent knows what to query
The main job of SQL Document Builder is to prepare content for an AI agent. Models can only work with what you give them: when every table, view, column, and routine is described in plain language—how it is used, what it means, and how it relates to the rest of the schema—the agent can narrow in on the correct objects instead of guessing from names alone. That makes thorough database documentation one of the most important steps in building a trustworthy natural-language-to-SQL workflow. Under the hood it is still the desktop SQL Server Script and Document Builder you rely on for deep object exploration, CREATE and data scripts, and up to 128 SQL tabs with saved connections and session restore. Built-in AI helps you produce and maintain those descriptions at scale (Describe missing, Describe with…, batch runs, smarter inheritance for views), and templates turn the result into Markdown, HTML, Wiki, SharePoint, or other formats your agent or team can consume—optionally persisting extended properties back to the database when you are satisfied. The application currently works with Microsoft SQL Server; support for other DBMS platforms is in development.
Template Switcher
## dbo.Customers
- Description: Stores customer records
- Primary Key: CustomerIDCore Features
AI that accelerates documentation
Use an on-box LLM to draft or refine descriptions, fill only undocumented fields, batch selected objects, and lean on view-definition parsing so inherited column context reduces noise and duplication before anything is sent to the model.
Explorer, scripts, and template output
Browse and filter the full object surface for Microsoft SQL Server, run scripts with safeguards, generate CREATE and data scripts in multi-tab editors—then ship consistent documentation through templates (Markdown, Wiki, SharePoint, JSON, and more) aligned with how your agent ingests context. Additional DBMS targets are on the roadmap.
Release Notes
v4.0.11
- Describe missing: fill only undocumented tables and columns while preserving existing text.
- View description intelligence: inherit column descriptions from base tables parsed from view definitions before LLM calls.
- Batch Describe: sequential processing with saves between objects and live progress.
- Search/replace: wrap-around fix, Replace All as a single undo step, Escape closes the panel without inserting text.