A2db

What is A2db?

A2db is an MCP server that enables batch SQL queries across multiple databases with read-only enforcement. It supports PostgreSQL, SQLite, MySQL, Oracle, and SQL Server, and allows pre-configured connections via .mcp.json. It is designed for developers and AI agents who need safe, multi-database querying from a single tool call.

How to use A2db?

Install with pip install a2db, then add to your MCP config using the command uvx with arguments to register one or more database connections. Alternatively, let the agent call login on demand — no pre-configuration is required.

Key features of A2db

• Batch queries with multiple named queries in one tool call
• Pre-configured database connections via .mcp.json
• Default connection for all queries in a batch
• Read-only enforced via SQLGlot AST parsing
• Clean JSON output with TSV data and per-query timing
• Supports 5 databases with all drivers bundled
• Error context with column suggestions and types

Use cases of A2db

• Running multiple analytical queries against different databases simultaneously
• Safely querying production databases from an AI assistant without write risk
• Simplifying data retrieval for reports across heterogeneous SQL databases
• Batch data extraction for ETL or validation pipelines

FAQ from A2db

What databases does A2db support?

PostgreSQL, SQLite, MySQL, Oracle, and SQL Server — all drivers are bundled with the package.

How does A2db ensure read-only queries?

It uses SQLGlot AST parsing to block write statements (INSERT, UPDATE, DELETE, etc.) at the syntax level before execution.

How do I configure connections?

Add entries to your mcpServers configuration with the register command and connection string, or let the agent call the login tool on demand without prior setup.

What output format does A2db return?

A JSON envelope containing TSV-formatted data rows and per-query timing information for each batch query.

Is there a default connection option?

Yes, you can set a default connection that applies to all queries in a batch, reducing the need to specify a connection per query.