Skip to content

Databases and Custom Tools

gary edited this page Aug 1, 2026 · 1 revision

Databases and Custom Tools

This area controls which user databases MCP keys can reach and which administrator-defined operations AI clients may invoke.

Database connections

DB Management supports SQLite, PostgreSQL, MySQL, SQL Server, Oracle, and Firebird.

  1. Name the connection and select its SQL provider.
  2. Enter the provider-specific connection values.
  3. Test connectivity.
  4. Save the definition and assign it to an MCP key.

Credentials are encrypted at rest. These user-database connections are separate from the internal Admin/Auth database.

Database management screen

Custom SQL tools

Use a custom tool when an operation should have an administrator-controlled structure instead of accepting arbitrary SQL from the AI client.

  1. Give the tool a clear name and description.
  2. Select query or DML behavior.
  3. Define its parameters and SQL definition.
  4. Parse, format, and test the definition.
  5. Save it and allow it on the appropriate MCP keys.

Custom SQL tool editor

Custom query tools inherit table, row, timeout, and concurrency enforcement. Custom DML tools also use the server's DML policy and human MCP Elicitation flow.

Semantic layer

Add human-readable table and column descriptions when database names alone do not convey business meaning. Schema discovery merges this metadata into the information returned to AI clients.

See MCP Tools Reference for runtime behavior and Security Governance for enforcement rules.

Clone this wiki locally