Dryrun is an MCP server tool from Boringsql, built with Go, that improves database development and operations. It connects to an offline snapshot of a PostgreSQL database, enabling AI agents, IDEs, and CI systems to gain schema awareness and perform validation. This approach allows for secure analysis without direct access to live production databases. Dryrun provides a suite of tools for schema exploration, query validation, and migration checks, supporting developers and DBAs in maintaining database health and consistency.

What does Dryrun connect to?

The Dryrun MCP server connects exclusively to an offline snapshot of a PostgreSQL database. This snapshot is stored locally within the project's ".dryrun/history.db" directory. The CLI tool is responsible for introspecting the PostgreSQL catalog and generating this local snapshot. This design means Dryrun does not establish direct connections to live production databases, ensuring that analysis is performed on a controlled, isolated copy of the schema and data structure, enhancing security and preventing unintended changes to operational systems.

Core features and tools

Dryrun exposes 14 tools for schema exploration, query validation, migration checks, linting, and vacuum health. Key functionalities include lock type analysis, duration estimates, table rewrite detection, and safe DDL alternatives. Query validation uses SQL parsing via libpg_query, column reference checks, and anti-pattern detection. Other tools offer snapshot diff, schema linting for conventions and structural audits. It also detects unused indexes, bloat, stale statistics, and activity anomalies.

Who is it for?

Dryrun is intended for AI agents, IDEs, CI systems, DBAs, and developers. It helps AI assistants understand database schemas, suggests indexes, and validates queries securely. DBAs and developers can use it for pre-deploy migration reviews, identifying disruptive operations, and enforcing schema governance. It supports maintaining database health through linting and statistical audits. Skip this if your workflow requires direct, real-time interaction with a live production database for analysis.

How to get started

No credentials are required for the AI assistant when connecting to the local Dryrun MCP server, as it operates on an offline snapshot. The server supports stdio or SSE as transport protocols.

Use cases

• AI-assisted database development: AI assistants gain schema awareness to suggest indexes and validate queries without production credentials.
• Pre-deploy migration review (Migration Guardrail): DDL statements run through lock and rewrite checks to identify disruptive operations and suggest safer alternatives before deployment or in CI.
• Schema Linting and Governance: Enforces naming conventions, identifies structural problems (e.g., missing primary keys, unindexed foreign keys), and detects statistical issues (e.g., unused indexes, bloat).

Dryrun offers a secure and isolated environment for PostgreSQL database analysis, particularly valuable for AI-assisted development and pre-deployment checks. Its reliance on offline snapshots means AI clients never access live database credentials, enhancing security. The tool suite supports a range of database health and governance tasks, from migration safety to schema linting.