Lessons Learned from Developing Agentic AI Tooling for Postgres
September 30 - October 2
AI agents are moving beyond writing code. They are querying data, taking actions, diagnosing performance issues, recommending changes, and performing work once reserved for experienced database engineers and administrators. But giving an agent access to Postgres, especially in production, raises a number of important questions.
Over the past twelve months pgEdge has released three major pieces of agentic AI focused tooling for Postgres, covering agentic engineering, agentic operations and agentic analytics. All of it is available for free with the Postgres open source license. Phillip will discuss design considerations and lessons learned by the pgEdge team. Some questions he plans to address include the guardrails needed to deploy MCP servers against both production and non-production databases, the extent to which organizations are ready to trust DBA activities to an agent versus having a human in the loop, and the need to give agents full access to historical data across live production databases and Iceberg data warehouses and data lakes.
This session offers a practical perspective on what it takes to make agentic AI genuinely useful and trustworthy across the Postgres lifecycle.
