The Beginning of Everything Introduces a Distributed Git Network for the Agent Era

Five months after attracting $60 million in seed funding at a cost of $300 million – called the largest seed round in the history of developer tools by its backer, Felicis – All is rethinking version control for agent-scale development, which allows developers to mirror GitHub repositories in an independent GitHub network built for agents.

The company, founded by former Microsoft GitHub CEO Thomas Dohmke and launched in February, built its own distributed Git network to address the limitations of standards, high latency, and outages of centralized platforms. It allows developers to manage repositories globally without a single central provider, Entire said in its announcement.

The preview, which opened under a waiting list as of today, enables developers to display an existing GitHub repository on the Yonke network, while the code stays where it is and agents “compile and pull from the regional view,” the company wrote. “This is a heavy load, read at once so agents can build without scale constraints.”

Looking ahead, data residency, governance and scale will be enabled when Core separates its network, so developers can host their own environments naturally.

According to Entire’s announcement, an end-to-end test of the reconstructed Git – the open source company – of the top agent function showed:

• Git Clone — how the repo is copied down for the developer or agent to run. It is fully supported ~570,000 clones/hour from a single repository.

• 200 simulated customers doing shallow-cloning from Frankfurt (40%), Paris, London, and Dublin in over 3 minutes.

• Git Push — how changes are sent to a shared repo, so that everyone (and every agent) can build on them. It is fully supported 586 pushes/second (or ~2.1 million/hour) at a single terminal or branch.

• 128 simulated agents push 1–10 files (2 KB each) per push over 2 minutes, each to its own branch. Tested in all native collections.

• Clone + Push (mixed) – a real-world loop to pull down the repo and push changes back, iteratively, the way the agent works. All configured ~470 clone + push operations/secondary operations in one repository.

• 128 simulation agents using shallow clone → 5 push → repeat, at ~50–60 ms delay p50.

“By design, Git was always intended to be distributed. As Linus Torvalds put it in his 2007 Google Tech Talk: ‘If you’re not distributed, you’re not worth using.’ In the era of agents, centralized Git hosting has become a significant obstacle, as the network of billions of agents and developers building a centralized server comes in the form of rate limitations, high latency, or outages,” Dohmke said in the announcement. “Today, we are starting to bring Git back to its original promise, with a distributed, and rapidly decentralized and fully open network of globally connected nodes. By doing so, we enable any developer or agent to host code locally, push, pull, and compile close to where it works, quickly and seamlessly, while still being part of the global network.”,

It all already provides a layer of semantic memory that includes all major coding agents, and helps agents stop repeating mistakes. It also allows developers to see how and why a piece of software is being developed. It automatically saves each session, quickly, the tool call and the test area directly in the repository next to the code.

Among its features are Complete Case, which shows why someone touched the code, including the agent session, immediately and the decision behind it; Update All, which sends a branch to multiple agents in parallel to receive a target-aware update; and code and semantic search, where developers and agents can search the history of code changes and why they were written.

“Time logs are now the second most important artifact in software development, and they rank next to code,” explained Dohmke. “With this layer of semantic memory tied to the repo, agents stop repeating mistakes, which means higher accuracy, more productivity, and lower token consumption. Developers can understand and verify what was built and why, which makes revisions much faster. It also opens up the opportunity to build a new developer life cycle, which allows us to understand and think about the large volumes that AI agents are now producing.”