An Entirely New Git Hosting Network<br>*:not([data-button-text])]:shrink-0 px-3 bg-neutral-900 dark:bg-neutral-100 text-white dark:text-neutral-900 hover:bg-neutral-800 dark:hover:bg-neutral-300 active:bg-neutral-700 dark:active:bg-neutral-400 focus-visible:[box-shadow:inset_0_0_0_4px_white] dark:focus-visible:[box-shadow:inset_0_0_0_4px_rgb(23_23_23)] font-medium site-pill-shadow hidden [html[data-has-session=true]_&]:flex">Dashboard*:not([data-button-text])]:shrink-0 px-3 bg-neutral-900 dark:bg-neutral-100 text-white dark:text-neutral-900 hover:bg-neutral-800 dark:hover:bg-neutral-300 active:bg-neutral-700 dark:active:bg-neutral-400 focus-visible:[box-shadow:inset_0_0_0_4px_white] dark:focus-visible:[box-shadow:inset_0_0_0_4px_rgb(23_23_23)] font-medium site-pill-shadow hidden [html[data-has-session=false]_&]:flex">Log in
*:first-child]:mt-0 [&>*:last-child]:mb-0">By design, Git was always meant to be decentralized. Every clone contains a complete copy of the repository and its history, allowing software to be replicated across many hosts rather than controlled by a single server. But in practice, Git hosting has routed developers into centralized platforms, built for a human-paced loop where a developer clones a repository once, works locally for a few hours, and makes the occasional push. This was sustainable until agents came along, sending thousands of concurrent requests in seconds, triggering traffic caps, and exposing failure points.
We believe that Git hosting must return to its original promise: a truly distributed network, not a system where the world’s software lives in a single location. And we built just that.
Starting today, you can mirror your public or private GitHub repositories on Entire—a fast, distributed, Git-compatible network. While your repo stays on GitHub, your coding agents fetch from a regional Entire cell that absorbs the heavy, concurrent read traffic, so they clone fast without hitting the origin repo’s rate limits. When you push a branch that is backed by GitHub, the push will be as fast as GitHub allows, but we also support Entire-native branches that can absorb much more concurrency.
We are launching with all existing Entire users and regions in the US, EU, and Australia. New users will be waitlisted, so we can monitor traffic and load. We’ll roll users off the waitlist progressively, aiming to reach full capacity in the coming weeks.
A Fast, Distributed Architecture Built for Agent Scale
Entire runs distributed nodes across multiple regions and jurisdictions, letting users pin their data to a single region or spread of multiple regions for higher redundancy and scale. In early testing, Entire’s Git network has shown strong results:
Cursor’s numbers are from their recent keynote. We measured our numbers with ForgeMark, a new open-source tool we’re releasing today under the MIT license. It benchmarks concurrent git-push throughput and latency against smart-HTTP forges under load patterns inspired by agent fleets. For workloads that need maximum throughput, any branch prefixed with entire/unmirrored/ stays local to the region it’s pushed to instead of syncing out to the mirror. That direct path is what powers the ForgeMark numbers above. It gives you a route that stays up independent of the mirror and handles high-volume write workloads at full speed.
Under the hood, Entire is a Git-compatible repository network with a global control plane for identity and placement, and regional data planes for content-addressed Git storage.
This design is optimized for:
ol]:mt-2 [&>ol]:mb-0 [&>ul]:mt-2 [&>ul]:mb-0 [&>ul]:list-[circle]">Scale: Entire scales Git traffic across regional API nodes and distributed object-storage providers. Ref updates remain protected by compare-and-set semantics, while object writes fan out across storage nodes, making the system well-suited to high-volume branch-parallel agent workloads.
ol]:mt-2 [&>ol]:mb-0 [&>ul]:mt-2 [&>ul]:mb-0 [&>ul]:list-[circle]">Low latency: Agents depend on fast repository reads: refs, commits, files, diffs, and merge bases. Entire serves those operations from the selected regional cell, keeping feedback loops close to the compute doing the work.
ol]:mt-2 [&>ol]:mb-0 [&>ul]:mt-2 [&>ul]:mb-0 [&>ul]:list-[circle]">Regional control: Repository content is stored and replicated in the region the user chooses. The global control plane coordinates identity, access, and placement, but does not become the central store for repository content.
ol]:mt-2 [&>ol]:mb-0 [&>ul]:mt-2 [&>ul]:mb-0 [&>ul]:list-[circle]">Availability: Writes are replicated across multiple availability zones. If a node leaves or falls behind, repair and catchup flows restore the replica set so repositories stay available without users having to move data manually.
In the coming months, we will decentralize and open source our Git network, we will allow for self-hosting, and we will rapidly expand our own network around the world. This will enable every...