As autonomous coding agents multiply, the infrastructure supporting Git repositories may face its biggest architectural rethink since the rise of cloud-hosted version control.
Git has long been the foundation of collaborative software development, built around a familiar workflow where developers clone repositories, make changes locally and periodically push updates to a central hosting service. But as AI agents take on a greater role in writing, testing and refining code, Karthik Rameshkumar, CTO of Entire, believes the infrastructure surrounding Git, not Git itself, may need a fundamental rethink.
In an exclusive interaction, Rameshkumar said the next bottleneck in AI-assisted software development may not be code generation itself, but the infrastructure that supports Git repositories. According to him, today’s Git hosting platforms were built for human developers, whose interactions with repositories are relatively predictable. AI agents, however, constantly analyse code, create branches, test changes and synchronise updates, often performing hundreds of repository operations while a developer might make only a handful of commits.
“As soon as you introduce AI agents into the development workflow, the nature of repository traffic changes,” he said. “You’re no longer optimising for people interacting with Git. You’re optimising for machines.”
Rather than replacing Git, Entire is building a distributed repository network that places synchronised repository mirrors closer to developers and AI agents. Instead of sending every repository request to a central hosting service, nearby mirrors handle repository operations while staying synchronised with one another.
According to the CTO, this architecture does not change the way developers collaborate on open source projects. Developers continue contributing to the same repository, while regional mirrors automatically stay in sync behind the scenes. This allows contributors in different parts of the world to work through the nearest repository mirror without affecting collaboration on a shared codebase.
He said lower latency becomes much more important when AI agents are involved. While a developer may clone or push code only occasionally, several AI agents working in parallel can repeatedly perform those same operations. If every request has to travel to a distant server, those delays quickly add up. Keeping repositories closer to where developers and AI agents are working reduces that waiting time while also easing the load on central hosting infrastructure.
The company also plans to open source key parts of its Git infrastructure in the coming months. Rameshkumar said Entire has re-engineered Git’s backend for distributed repository management and plans to release core components, including what it calls EntireDB, to encourage developers to build on the technology while continuing to operate its hosted platform.
Whether that vision gains widespread adoption remains to be seen. But as AI agents move from coding assistants to active engineering collaborators, Rameshkumar believes the infrastructure behind Git will also need to evolve.
The broader discussion also explores how AI-native developer platforms could reshape version control, software governance and engineering collaboration as AI becomes a more active participant in the software development lifecycle.
