Dory

Dory is a desktop data workspace that brings a SQL editor, schema explorer, and AI copilot together without requiring you to switch between tabs or tools. It's built for analysts and developers who work with databases daily and want AI assistance that stays honest about what's actually in their schema.

The core problem it solves: most AI SQL tools hallucinate column names and table structures. Dory grounds every suggestion in your real schema. The AI copilot reads your actual tables, fields, and relationships before generating SQL or explaining results, so you get queries that run rather than ones you have to debug.

Key capabilities:

• Schema-aware SQL generation reads your live schema before writing a single line, so autocomplete and AI suggestions reflect real columns and joins
• SQL editor with keyboard-first editing, multi-tab workflows, and high-performance result grids with sorting and filtering
• Schema Explorer lets you browse tables, columns, and types in the same context the AI uses
• Saved queries and results turn one-off exploration into reusable workflows you can revisit and share
• Database chat lets you ask questions in plain language and get real SQL back, not generic responses
• Auto table intelligence generates semantic annotations and usage cues so the AI understands business data before writing queries
• SSH tunneling encrypts every connection so credentials and query data stay within your network

Database support covers PostgreSQL, MySQL, SQL Server, Oracle, SQLite, MariaDB, DuckDB, and ClickHouse, with a deeper integration for ClickHouse including query analytics dashboards and team-ready workspace features.

Dory follows a local-first philosophy. Connections, personal settings, and workspace metadata stay on your machine by default. It's Apache 2.0 licensed and self-hostable via Docker. A free tier covers personal projects and local workflows.

For teams already using tools like Chat2DB or notebook-style environments like Deepnote, Dory positions itself as the option that keeps AI grounded in real schema context rather than relying on general language model knowledge.