GemDB launches Code extension for AI-driven schema changes

GemDB has launched GemDB Code, a free Visual Studio Code extension for developers and AI coding agents that need to change application data models.

The extension packages GemDB's object database engine inside Visual Studio Code and uses application objects as the database schema. Developers or AI tools can create or modify an object in code and persist it on commit without a separate migration step.

The approach addresses a longstanding software development problem: changes to application code often require parallel updates to a relational database schema. GemDB argues that AI-generated code is increasing pressure on that process because database changes still need manual coordination, migration scripts and maintenance windows, even when application logic can be produced much faster.

Brad Shimmin, Vice President and Practise Lead, Data Intelligence, Analytics, and Infrastructure at Futurum, said the timing reflects a shift in the object database market.

"The object database category has been largely dormant for a decade, but the conditions that held it back have changed," said Brad Shimmin, Vice President and Practise Lead, Data Intelligence, Analytics, and Infrastructure at Futurum. "AI coding agents are exposing the schema migration bottleneck in ways that were theoretical even two years ago. GemDB is entering this moment with an in-process object database that provides full ACID guarantees and a production track record in financial services spanning three decades."

GemDB Code is available as a free download and includes a scale-limited version of the database engine for local development. Python is supported at launch.

How it works

GemDB's premise is that objects persist in the same form they take in application code. In practice, altering a class or object definition changes the schema at the same time, rather than requiring separate mapping layers, join tables or migration scripts.

That can simplify tasks such as adding many-to-many relationships, which in a relational database would typically require new tables, foreign keys and data backfills. In GemDB Code, the same change can be made by adding a field to an object.

Sandeep Parikh, Chief Product Officer at GemDB, said the release is aimed at AI-assisted development workflows.

"We've spent decades accepting a separation between code and data that was not inevitable," said Sandeep Parikh, Chief Product Officer at GemDB. "Any builder working with a coding agent eventually hits the same wall, where the agent writes application logic in minutes, but the database layer still expects manual coordination, migration scripts, and scheduled downtime. GemDB removes that wall entirely. When your objects are the database's objects, schema changes become code changes, and code changes are what AI agents do best."

The extension also includes a built-in Jupyter Notebook kernel for working against a live database and an MCP server for natural-language queries through chat-based AI agents. GemDB said this allows developers to interact with data conversationally rather than through formal database queries.

Industry roots

GemDB's underlying technology is based on GemStone/S, which it said has been in continuous production since 1986. The company added that the database has been used in regulated sectors including financial services, with deployments at organisations such as JPMorgan Chase and UBS.

GemDB was spun out of VMware 13 years ago. Emergence acquired the company in April 2026, giving it backing to develop new language runtimes and AI-focused features while maintaining compatibility for existing users, according to GemDB.

GemDB also used the launch to highlight customer experience in insurance. Lifeware, a database user, said the object-based model removed the need to reshape business processes into rows and columns.

"Everything you can think of in a life insurance company that we could digitize and automate we put into GemDB," said Massimo Arnoldi, Founder and Member of the Board at Lifeware. "Policy administration, claims processing, underwriting workflows, agent compensation. All of it runs on objects that map directly to how our business actually works. We never had to flatten a life insurance contract into rows and columns, so our developers spent their time building insurance logic, not fighting with the database. When we need to change a product or add a rider, the code change is the only change."

Market question

The release comes as software companies look for ways to remove bottlenecks created by the rise of AI-assisted coding. While AI tools can generate application logic quickly, businesses still need a way to move those changes into testing and production without adding operational risk.

GemDB argues that the database layer is now a limiting factor, especially when teams must coordinate schema updates separately from application releases. Its answer is not another synchronisation tool or abstraction layer, but a development model in which the schema is embedded directly in the codebase.

Kit Merker, Chief Executive Officer at GemDB and Operating Partner at Emergence, said that distinction is central to the company's pitch.

"AI coding agents are changing application code faster than any team can coordinate database updates to keep pace," said Kit Merker, Chief Executive Officer at GemDB and Operating Partner at Emergence. "Other approaches only give you better coordination. GemDB eliminates this need altogether by ensuring that when your agent changes the code, it has already changed the schema. There is no second step."