How to Set Up SaturnSQL Context
Last updated November 19, 2026 · By the SaturnSQL team
SaturnSQL Context is what AI tools read about a connection before they write SQL: context files, business definitions, and what each table and column means. There are four ways to fill it in, and you can mix them.
1. Write it in SaturnSQL
Open Context in the app and pick the connection in the header. Everything saves as you type.
- Context files: Markdown files in folders with rules that apply to every question, such as “amounts are in cents” or “use orders_v2, not orders”. Each file and folder can have an owner, and access can be set per group.
- Definitions: business terms like “active customer”, with the SQL that computes them.
- Tables and columns: what each one holds, other names people use for it, and who owns it, so AI tools can point people to the right person or team. Your text replaces the database’s own comment for AI tools; where you write nothing, the comment shows through.
Start with the mistakes you have already seen an AI make. Each one is usually a line in a context file.
2. Let an AI tool draft it
In Context, choose Set up with AI and copy the prompt into Claude Code, Cursor or Claude. The tool reads your code and data through SaturnSQL’s MCP connector, drafts context files, definitions and descriptions, and checks with you before it saves anything. Every change is recorded with your name in Recent changes.
Saving needs the connection owner or an admin, and an API key or sign-in with the write scope.
3. Import an Apache Ossie model
If your team already has a semantic model in Apache Ossie, choose Import in Context and pick the YAML file. SaturnSQL shows what will change before it saves: descriptions, synonyms, context files and definitions, plus anything in the file the database doesn’t have. Nothing the file leaves out is removed. How Apache Ossie import and export work.
4. Keep it in git
Export the connection as Apache Ossie YAML, commit it next to your code, and a GitHub Actions workflow previews each change on the pull request and updates Context on merge. Use a Context sync only API key, which can import and export context and nothing else. Keep AI context up to date with GitHub Actions.
Check what AI tools see
Ask your AI tool a question you know the answer to, such as last month’s revenue, and compare. When it gets one wrong, add the context that would have prevented it. The Overview in Context shows how much is written for each connection.
New to the idea? Read what a context layer is.