When AI Agents Write Your Code, Who Protects Your Database?
No video of the event yet, sorry!
Legacy dev workflows (nightly dumps, hand-crafted seed data, shared staging environments) are already strained. AI coding agents make it worse: they generate migrations and queries faster than any team can review, they experiment constantly, and they need fresh data and instant resets to work effectively. The database becomes the bottleneck.
This talk demonstrates a practical workflow for giving coding agents safe, isolated access to production-representative PostgreSQL data. Using a live demo, we walk through an MCP-integrated loop: spinning up instant anonymized clones, letting an agent work against real data distributions, and catching problems (like a migration that fails on production edge cases) before they ever reach production. We'll also show how we teach an old dog a new trick: git-like branching for your database, without modifying a single line of Postgres.
- Date:
- 2026 April 23 09:00 PDT
- Duration:
- 50 min
- Room:
- Winchester 2
- Conference:
- Postgres Conference: 2026
- Language:
- Track:
- Dev
- Difficulty:
- Medium