Queryshield
Route agent-generated natural language into audited, row-level-scoped SQL instead of handing your database credentials to the model.
Overview
QueryShield is a MCP server for the Operate phase that proxies agent SQL with NL→SQL, AST safety, per-agent RLS, and audit logging.
What is this MCP server?
- Natural language to SQL with AST-based safety checks before execution
- Per-agent row-level security (RLS) so each agent sees only allowed rows
- Audit log of agent queries for production accountability
- PyPI stdio package queryshield-mcp with QUERYSHIELD_API_KEY
- Optional QUERYSHIELD_BASE_URL override for non-default deployments
- Server version 1.0.1 via PyPI package queryshield-mcp
- Capabilities: NL→SQL, AST safety, per-agent RLS, audit log
- Optional QUERYSHIELD_BASE_URL endpoint override
Community signal: 1 GitHub stars.
What problem does it solve?
You want your coding agent to query the database, but direct credentials and unreviewed SQL are an unacceptable leak and blast-radius risk.
Who is it for?
Indie operators running agent-assisted internal tools or support bots who need NL queries with RLS and auditability on a real database.
Skip if: Greenfield projects with no database yet, or teams fine with read-only static exports instead of live governed SQL.
What do I get? / Deliverables
After you issue agent keys and connect the MCP, queries flow through QueryShield with safety checks, scoped rows, and logs you can review.
- Agent-executed SQL passing AST safety and RLS boundaries
- Audit log entries for accountability on agent-driven queries
Recommended MCP Servers
Journey fit
A secure SQL proxy matters when agents touch production or staging data—a post-ship operations and infra control. Infra is the shelf because QueryShield sits between agents and your database with policy, AST checks, and logging.
How it compares
Secure database proxy MCP—not a generic ORM code-generation skill.
Common Questions / FAQ
Who is QueryShield for?
Builders who already expose databases to AI agents and need AST-checked SQL, per-agent RLS, and audit logs instead of sharing connection strings.
When should I use QueryShield?
Use it in Operate when agents run recurring analytics or ops queries against staging or production and you must enforce scope and traceability.
How do I add QueryShield to my agent?
Install queryshield-mcp from PyPI, set QUERYSHIELD_API_KEY (and QUERYSHIELD_BASE_URL if needed), register the stdio server in your MCP client, and restart.