Key Takeaways
- Entire launches as a distributed Git alternative built specifically for AI coding agents
- The platform claims 25x better performance than competitors like Cursor Origin for agent-heavy workloads
- GitHub's frequent outages from AI coding traffic created the opening for this new infrastructure play
Thomas Dohmke, GitHub's former CEO, launched Entire on Wednesday: a decentralized Git network designed to handle the crushing load that AI coding agents place on traditional version control. The platform enters preview with servers in the US, EU, and Australia, offering developers an alternative to GitHub's increasingly strained infrastructure.
Why build a new Git network now?
GitHub has suffered repeated outages in recent months as vibe coding exploded and autonomous coding agents flooded its centralized servers. When GitHub goes down, developers worldwide stop working. The platform is trying to scale up, but Dohmke apparently decided the problem needs a structural solution, not just more capacity.
Entire distributes repositories across multiple geographic hubs. The architecture spreads risk and cuts latency for developers tired of depending on a single point of failure. In the preview, users can mirror an existing GitHub repository to Entire in one step. Code stays on GitHub while agents clone and pull from regional Entire mirrors, offloading the heavy concurrent reads that trigger rate limits.
Performance claims: 25x faster than competitors
Entire published initial benchmarks. The platform handled roughly 570,000 clones per hour from a single repository during testing. That test simulated 200 clients shallow-cloning from European cities, including Frankfurt, London, Paris, and Dublin, completing in about 3 minutes.
For pushes, Entire maintained 586 per second to a single branch, which works out to about 2.1 million pushes per hour. The company compared this to Cursor Origin, another recently announced agent-first platform, which claimed 81,360 pushes per hour in a keynote. Entire says it benchmarks 25x ahead of competitors' claims.
The platform works with Codex, Claude Code, Cursor, GitHub Copilot, and other popular coding agents. It stores session data alongside code, which matters when you need to trace what an agent did and why.
Features designed for agent supervision
Beyond raw speed, Entire targets a pain point specific to agentic coding: knowing what the agents actually wrote. The company describes a "semantic memory layer" that catches agent mistakes and gives developers visibility into AI-generated code. Two features stand out here.
Blame surfaces who last edited a line of code and why. Review prompts multiple agents to reflect on an action before committing it. Both address the growing problem of AI agents making changes that humans never explicitly approved.
The company plans to let programmers host new public and private repositories natively, moving beyond the mirror-only model. It also intends to fully decentralize the network, enabling data residency and sovereignty in any region. That last point matters for enterprises bound by local data regulations.
The risk of building on venture-backed infrastructure
A note of caution is warranted here. Entire is a venture-backed startup. Services with external funding can lose that funding and shut down without much warning. Developers who build critical workflows on top of Entire accept that risk.
GitHub, for comparison, has Microsoft's resources behind it. That backing does not prevent outages, but it does reduce the chance that the service disappears entirely. Entire's decentralized architecture might actually mitigate some shutdown risks, since data would be distributed rather than centralized. But the company's long-term viability remains unproven.
Developers can join the waitlist now. The preview is open to anyone, with servers available in the US, EU, and Australia. Users can mirror into multiple regions simultaneously.
Logicity's Take
Entire's timing is sharp. GitHub's outages during peak AI coding periods created real frustration, and Dohmke knows that frustration intimately from his CEO tenure. The 25x performance claim needs independent verification, but even if the actual advantage is smaller, the decentralized model addresses a structural weakness in how developers currently work. The interesting question is whether GitHub responds by building similar distributed infrastructure, or whether Microsoft views this as a niche problem. For engineering teams evaluating Entire, the practical move is to use it as a read mirror first. Let agents pull from Entire while keeping GitHub as the source of truth. That setup captures the performance benefits without betting the codebase on a new platform.
Frequently Asked Questions
What is Entire and who founded it?
Entire is a decentralized Git network built for AI coding agents. It was founded by Thomas Dohmke, who served as GitHub's CEO from 2018 to 2021.
How does Entire compare to GitHub?
Entire uses a distributed architecture with servers across multiple regions, while GitHub relies on centralized US-based servers. Entire claims 25x faster performance for agent-heavy workloads.
Which AI coding tools work with Entire?
Entire supports Codex, Claude Code, Cursor, GitHub Copilot, and other popular coding agents. It also stores session data alongside code for better agent tracking.
Is Entire available now?
Entire launched in preview on July 8, 2026. Developers can join the waitlist and access servers in the US, EU, and Australia.
Can I keep using GitHub with Entire?
Yes. In the current preview, Entire mirrors existing GitHub repositories. Your code stays on GitHub while agents clone from regional Entire mirrors.
Another major AI infrastructure announcement this week
Need Help Implementing This?
If your engineering team is evaluating distributed Git infrastructure or managing AI coding agents at scale, reach out to Logicity's consulting partners for architecture guidance tailored to your stack.
Source: Latest news
Huma Shazia
Senior AI & Tech Writer
Produced with AI assistance and reviewed by the Logicity editorial team. Learn more in our Editorial Policy.
