Adronite launches Codistry AI coding platform, claims half the token cost
Adronite Inc. today launched Codistry, an artificial intelligence coding platform for large enterprise codebases. The platform runs on Adronite’s Context Engine, or ACE, which the company has patented. ACE builds a relational map of a codebase and keeps it current as the code changes. A model working on a task gets only the parts of […] The post Adronite launches Codistry AI coding platform,…
Adronite Inc. has introduced Codistry, an artificial intelligence coding platform designed for large enterprise codebases. This platform operates on the company's patented Context Engine, or ACE. ACE constructs a relational map of the codebase, which remains up-to-date as the code evolves. ACE utilizes only the specific portions of this map that are required for a given task.
The indexing process begins upon installation and does not necessitate extensive preparation beforehand. Adronite asserts that shipping extensive codebase portions as prompts with each request contributes to high token costs. ACE maps the codebase once and reuses this map, resulting in smaller prompts as teams employ AI tools more frequently on the same code.
According to the company, benchmark tests indicate that Codistry consumes approximately half the tokens compared to Anthropic PBC's Claude Code for similar development tasks. Both tools operate using Claude Opus 4.8 in the cloud. The average cost per task is about 48% lower using Codistry. In an open-source backend project, PocketBase, the per-task cost dropped from $2.12 to $1.10 when using Codistry, inclusive of the one-time indexing cost divided across subsequent tasks.
Adronite's CEO, William Colleran, highlighted that many AI coding tools fall short due to a lack of necessary context to comprehend how a codebase operates. He emphasized that organizations should not have to compromise between expensive frontier models, safeguarding their intellectual property, or obtaining high-quality AI assistance.
CTO Edward Rothschild explained that ACE provides only the context required for a task at the time it is needed, allowing the model to devote more of its reasoning capacity to the problem at hand. Codistry is compatible with frontier models, and customers can also integrate open-weight models they host themselves. Deployment options include public and private clouds, on-premises servers, and air-gapped environments, ensuring that source code remains within the customer's infrastructure.
The company is targeting regulated industries and midmarket companies, many of which cannot send proprietary code to an external endpoint. Adronite is currently hosting a 72-hour developer challenge to promote the launch, where participants build an interactive web app on Codistry under a fixed token budget. The first prize is $5,000, with runners-up receiving developer-configured Mac minis.
Rothschild co-founded Adronite in Seattle in 2023 and transferred leadership to Colleran, previously the CEO of RFID chipmaker Impinj Inc., in July. Adronite raised $5 million in a Series A round in February led by Gatemore Capital Management.
Written by urgent.news from SiliconANGLE's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.