Adronite today launched an artificial intelligence (AI) coding tool based on a context engine that understands relationships within an existing codebase.
Dr. William T. Colleran, CEO of Adronite, said the Codistry AI coding tool leverages an Adronite Context Engine (ACE) to surface context in a way that generates better code at a significantly lower total cost because fewer tokens are consumed much more efficiently.
Before any code is generated, ACE first maps the underlying software architecture, dependencies, vulnerabilities and relationships within a codebase. Those insights are then used by Codistry to better understand how the code being generated will impact a codebase, said Colleran.
Designed to work with self-hosted open-weight AI models, that approach eliminates the need to share large portions of a codebase to AI models, which in turn reduces the number of tokens that might otherwise be consumed, he added. That codebase can also be based on multiple programming languages that ACE is able to map, noted Colleran.
Adronite claims that its internal benchmarking against Claude Code running on the Opus 4.8 AI model, Codistry completed comparable development tasks using roughly half the tokens to rescue costs by nearly half (48%). On the PocketBase open-source codebase, Codistry reduced per-task costs from $2.12 to $1.10, the company claims.
ACE was originally designed to provide insights for a documentation tool, but with the rise of AI it became apparent that the same context could be used to make coding tools more efficient, said Colleran. The end result is an AI coding tool that enables a developer to more easily express an intent without having to deeply understand the underlying codebase to generate code that can actually be used in a production environment, he added.
Mitch Ashley, vice president and practice lead for software lifecycle engineering at The Futurum Group, said that context is critical because developers now spend more time explaining gaps in context and understanding across a codebase to an AI model than reviewing its output. Adronite’s context engine maps dependencies before it generates code in a way that benefits the code review process, he added.
While adoption of AI coding tools is now widespread, it’s not clear how much of the code being generated using these tools is actually making it into production environments. Additionally, the quality of that code is often suspect, which has led to a general increase in the number of issues that DevOps teams need to address after an application has been deployed.
Regardless of approach, the one thing that is clear is that AI models are rapidly becoming interchangeable commodities. If the cost of using one rises or another becomes more advanced, a harness such as Codistry significantly reduces the cost of switching from one model to another. In fact, some developers, depending on their sensitivity to cost, will assign the same task to multiple AI models to determine which one generates better output.
The challenge, as always, is making sure that however code is created, it is worth more than any trouble it might cause further down the DevOps pipeline.