MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202621110575 A) filed by Samruddhi Nichal; Hrishikesh Vallakati; Ananya Borkar; and Dr. Amol B. Kasture on September 15, 2026, for Forgenta Ai: System And Method For Multi-Agent Software Engineering And Validation.
Inventors include Samruddhi Nichal; Hrishikesh Vallakati; Ananya Borkar; and Dr. Amol B. Kasture.
The application for the patent was published on October 30, 2026, under issue no. 44/2026.
Abstract: Anyone who has spent real time on a real codebase knows that writing the code is usually not what eats up the day. It is everything that happens around the code: running the linter, waiting on a security scan, tracking down why a test broke overnight, writing documentation nobody particularly enjoys writing, and then pulling the results of all that together before anything is actually ready to ship. Each of these tools does its own job reasonably well on its own, but they do not talk to each other. That leaves the developer as the glue — reading one tool's output, deciding what actually matters, turning it into a fix, and then running the next tool to see if the fix worked. Do that enough times over the life of a project and the coordination itself becomes the bottleneck, not the coding. Forgenta AI is proposed as an autonomous AI operating system for software engineering, aimed less at speeding up that coordination work and more at removing it. Rather than being another chat-based assistant that sits waiting for someone to type a prompt, it is designed as a multi-agent system that behaves more like a small engineering team of its own. It takes in a software repository, builds up a structured understanding of how the frontend, backend, database, and APIs actually connect, figures out what needs attention, and routes that work to a set of specialized AI agents — one handles debugging, another security, another testing, another performance, and so on — that carry it out with minimal supervision. An orchestration layer sits underneath all of this, coordinating the agents, managing dependencies between tasks, retrying what fails, and keeping a record of exactly what happened at every step so the process stays auditable instead of turning into a black box. Before deployment is allowed, a validation gate checks security and testing results against predefined conditions, and anything destructive or production-affecting still needs a human to sign off. Put together, the system combines large language models, multi-agent orchestration, static analysis, automated testing, knowledge graphs, databases, containerization, and CI/CD tooling into one integrated platform for software engineering automation.
Disclaimer: Curated by HT Syndication.
