AI Coding Agents Fuel New Demand for PostgreSQL, pgEdge CEO Says

"It's the default database for agent-generated applications," Merrick said. "Demand has skyrocketed."

Merrick's assessment follows an estimate from Databricks Inc. that AI agents now spin up 80% of enterprise databases. PostgreSQL gives those agents a common foundation for building applications, retrieving information for language models and helping administrators manage the resulting database sprawl, according to SiliconANGLE.

AI has also pushed new kinds of database features into the development mix, including vectors, graphs and agentic branching. Merrick said PostgreSQL's reputation for stability and performance, combined with an extensible architecture, appeals to developers who are otherwise overwhelmed by new database architectures.

That extensibility lets developers add AI capabilities without adopting a separate database engine. The pgvector extension, for example, stores vectors alongside other data and supports similarity searches that underpin retrieval-augmented generation. Because PostgreSQL's source code and documentation are publicly available, coding agents can also read them and use them to build more applications.

Keeping application records and vector representations inside PostgreSQL reduces the number of database technologies a team must learn and operate, and it gives retrieval workloads access to established database features such as transactions and joins.

Merrick said the growth in agentic application development is multiplying the number of databases that must be managed, while the supply of human database administrators stays fixed.

pgEdge is among the vendors building a business on the need to balance rapid experimentation with the security, operational control and deployment flexibility enterprises require. The company recently announced its Starfleet platform, aimed at infrastructure that lets an application be built easily and supported reliably in production.

Merrick argues that standardizing on fewer technologies lowers training and management overhead, though it does not reduce the workload itself. That opens a second role for AI: helping administrators diagnose problems and propose fixes.

pgEdge's AI DBA Workbench monitors databases, identifies issues and generates SQL to resolve them, subject to approval by a human administrator. Merrick described a case in which a report's runtime suddenly stretched from minutes to hours; an agent might trace the cause to a configuration change or a missing index, but investigating a problem should not grant it unrestricted authority to alter the database. pgEdge's SafeSession technology enforces read-only access at the PostgreSQL engine level, and its Model Context Protocol server connects agents to database contents and schema information so that access controls are part of the integration itself.

According to pgEdge, customers can build prototypes with Starfleet in the company's cloud and then deploy the same PostgreSQL distribution in a managed cloud, a customer-controlled cloud or on-premises. The company says that flexibility addresses cases where data location rules or air-gapped infrastructure constrain production choices, and it accommodates the growing popularity of open-source and open-weight databases that businesses want to run on-premises or in a private cloud.

Merrick said PostgreSQL is becoming a meeting point for coding, retrieval and operations, supported by a broad developer base. He cited a saying common among developers: for just about any data management problem, the answer is Postgres.

Editor's Summary

PostgreSQL again led Stack Overflow's developer survey, and pgEdge chief executive Phillip Merrick credits AI coding agents for renewed demand, saying the engine is the default for agent-generated applications. pgEdge is positioning its Starfleet platform and AI DBA Workbench around that trend, pairing agent-driven database management with human-approved, read-only safeguards. The company's pitch reflects a broader shift in which coding, retrieval and database operations increasingly converge on a single engine.