Connect AI to governed data operations

AI and SQL MCP

AI can accelerate SQL development and help users work with data, but it needs context, review, and a controlled execution boundary. Keep development assistance inside a reviewable workflow and expose named database operations instead of unrestricted credentials or raw SQL access.

Core focus

What this practice covers

AI-assisted development

Ground prompts in schema and repository rules, inspect proposed changes, and validate output against development data.

Governed tool access

Expose approved tables, views, procedures, and actions through Data API Builder or narrowly scoped MCP tools.

Security and evaluation

Protect secrets, separate permissions, log tool use, test denied operations, and measure correctness before deployment.

Typical work

What teams work on

Copilot repository instructions and SQL review skills
Prompt patterns for SQL generation, review, and troubleshooting
Data API Builder REST, GraphQL, and MCP configuration
Entity roles, field restrictions, and row policies
Tool descriptions and database operation design
Evaluation of SQL correctness, safety, and performance

Results

What good looks like

More consistent AI-assisted SQL changes
No raw production credentials in prompts or clients
A narrow and auditable data-access contract
Human review and database evidence retained in the workflow