Adronite launched Codistry, an AI coding platform that uses a patented Context Engine to map codebases and feed AI models only the code context they need for each task.
In benchmarks, the platform cut token costs roughly in half compared to Anthropic's Claude Code—bringing per-task cost down to about 48% lower—by avoiding the need to repeatedly ship large code chunks into prompts.
The platform is designed for enterprise teams, regulated industries, and companies that require on-premises or air-gapped deployment to keep proprietary code off external servers.
What happened
Adronite Inc. launched Codistry, an AI coding platform for large enterprise codebases, powered by its patented Context Engine (ACE). ACE builds a relational map of a codebase and delivers only the context each task requires, rather than sending entire code chunks into prompts on every request.
Why it matters
In Adronite's own benchmarks, Codistry used roughly half the tokens of Anthropic's Claude Code on comparable tasks, with average cost per task about 48% lower. On the open-source PocketBase project, per-task cost fell from $2.12 to $1.10. For regulated industries and midmarket companies that cannot send proprietary code to external endpoints, the platform offers deployment across public and private cloud, on-premises, and air-gapped environments, keeping source code within customer-controlled infrastructure.
What to watch
Adronite is running a 72-hour developer challenge tied to the launch, where entrants work within a fixed token budget to build an interactive web app; first prize is $5,000. The company raised $5 million in Series A funding in February, led by Gatemore Capital Management.
Ask the AI about this article →
Adronite's entry into the AI coding space addresses a core cost and context problem: feeding large code chunks into prompts on every request inflates token usage and expenses. The company's patented Context Engine takes a different approach—it builds a relational map of the codebase once during installation and reuses that map, supplying only the context each task requires. This strategy aligns with the framing offered by Chief Technology Officer Edward Rothschild: most AI coding platforms solve hard engineering problems by "feeding increasingly powerful models more context and tokens," whereas ACE supplies only what is needed when it is needed, leaving more of the model's reasoning capacity for the actual problem.
The benchmarks Adronite published—showing roughly half the token cost and 48% lower average cost per task compared to Claude Code, with the PocketBase case dropping per-task cost from $2.12 to $1.10—are compelling for cost-conscious enterprises, especially those in regulated industries or with strict intellectual-property controls. The company's targeting of regulated industries and midmarket firms, combined with support for on-premises and air-gapped deployment, reflects a deliberate focus on customers who cannot send proprietary code to external endpoints. This positioning opens a potential market segment beyond cloud-native startups.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
Slack introduced Slack Code, a new feature that lets teams collaborate with AI coding agents (Claude, Devin, G…
Citi, HSBC, and Standard Chartered have adopted Ant International's artificial intelligence tool for foreign e…
SpaceX closed a $60 billion acquisition of Cursor, a popular code editor with over 50,000 companies in its use…
Adobe announced general availability of audio generation capabilities in Firefly, its creative AI suite
AWS Professional Services built a multi-agent framework running on Amazon Bedrock AgentCore that automates ent…
Slack has introduced Slack Code, a feature that lets AI coding agents work in dedicated project channels where…
