An AI code-intelligence company teaching machines to read the decades-old software still running banks, airlines, and government - and handing the translation to the humans and AI agents who have to maintain it.
It is an ordinary Tuesday inside an ordinary bank. A card is swiped in a coffee shop, an approval flashes back in under a second, and nobody thinks twice. Underneath that second sits a program written in a language most universities stopped teaching before the maintainers were born. No one in the building can fully explain it. The person who could has long since retired, taking the mental map with them. This is the quiet emergency CoreStory was built for - not a flashy frontier, but the basement of modern civilization, where the lights still work and no one has the wiring diagram.
CoreStory's founder, Anand Kulkarni, likes to put it plainly: "Every time you swipe a card or make a claim, you're depending on software written decades ago that few people understand." The trillions of lines of legacy code that run Social Security, Medicare, airline reservations, insurance claims, and defense networks did not vanish when the cloud arrived. They simply went undocumented, accruing a kind of debt that compounds in silence. CoreStory's bet is unfashionably sensible: before you can safely change something, you have to understand it. So it built a machine that reads first and rewrites never.
Every time you swipe a card or make a claim, you're depending on software written decades ago that few people understand.
Point CoreStory at a sprawling, undocumented codebase and it does not panic. It ingests the source, builds a persistent intelligence model of how the whole thing actually works, and then keeps that understanding alive - serving accurate context on demand to whoever asks. The flagship, Code-to-Spec, reads millions of lines and writes them back out as living requirements: feature overviews in plain English, architecture diagrams, annotated explainers, dependency maps, and modernization recommendations. The team calls the result "an atlas for code." It is a good name, because the point of an atlas is not to admire it - it is to go somewhere.
Reads legacy code and generates living specs: natural-language feature overviews, architectural summaries, visual code-snippet explainers, dependency maps, and modernization recommendations - with a chatbot to query it all.
Ingests source and builds a durable model of the system that serves accurate context on demand - to architects, product managers, developers, and AI coding agents alike.
Lets AI agents - Claude Code, GitHub Copilot, Cursor, Devin, Droid - query CoreStory's intelligence directly, so the agent works from the map instead of guessing.
Guided spec creation for common workflows, plus ready-to-use playbooks with pre-built integrations for the leading coding agents.
CoreStory is aimed squarely at the enterprise basement: banks, airlines, insurers, and government-scale systems carrying long-lived codebases in everything from COBOL to TypeScript. The new hire who needs to ship in week one. The architect mapping a migration. The product manager trying to learn what a forty-year-old module even promises to do. And, increasingly, the AI coding agents now expected to modernize systems they have never seen. To all of them, CoreStory hands the same thing: a shared, queryable understanding - so the answer to "what does this do?" stops being a séance for a retired engineer.
In October 2025, CoreStory closed a $32M Series A. The thesis is not subtle: AI can now write code faster than ever, which only sharpens the older problem of understanding the code already there. The round was led by a trio of believers, with a notable bench behind them.
Lead investors: Tribeca Venture Partners, NEA, SineWave Ventures.
CoreStory is redefining how enterprises understand and manage their software estates.
An applied-math and AI builder who previously co-founded LeadGenius and created CrowdMath. He frames CoreStory's mission as making the world's software infrastructure understandable again. LinkedIn
Co-founder helping build the model ensembles and agentic workflows behind CoreStory's code-intelligence engine.
CoreStory enters the market with its AI code-intelligence platform, Code-to-Spec, for automatically documenting legacy codebases.
Raises a $32M round led by Tribeca, NEA and SineWave to scale code intelligence for enterprise modernization.
Joint research shows structured CoreStory specs improve AI software-engineering agent accuracy by 51%.
Product walkthroughs and talks live on CoreStory's channel and site. Start here.
The card swipes. The approval flashes back in under a second, same as ever. Nothing visible has changed - which is exactly the point. But somewhere in the bank's basement, the program that cleared it is no longer a black box. There is a living spec describing what it promises, a diagram of how it connects, and an AI agent that can answer "what does this do?" without summoning a ghost. The retired engineer's mental map has been rebuilt and made shareable. CoreStory did not make the old code disappear. It made it legible - and in a world quietly held together by software nobody remembers writing, legible is the whole game.
BusinessWire (Series A) · SiliconANGLE · Techzine · Intelligence360 · Pulse 2.0 · FinSMEs · corestory.ai