Dev.to
7/18/2026

Why I built a SQL client when 10 already exist — then let AI agents into it
Short summary
A developer built a keyboard-first SQL client called data-peek that exposes a local MCP server for AI agents to query databases safely. Read-only queries run inside rolled-back transactions enforced at the database level, while writes require human approval via an in-app dialog. All statements are logged to a tamper-evident hash chain, giving developers control over what agents actually execute.
- •data-peek runs a local MCP server so AI agents like Claude Code can query databases with guardrails
- •Read-only enforcement uses database-level transaction rollback, not SQL string matching; writes require explicit human approval
- •All agent and human queries land in a hash-chained audit log that detects tampering
Generated with AI, which can make mistakes.
Is this a good recommendation for you?



