Entire Distributed Git Network: A Developer Guide to the Ex-GitHub CEO's Agent-Era Platform

Last updated: July 14, 2026

Thomas Dohmke left GitHub, raised $60 million, and is now building a distributed Git network designed for the demands of AI coding agents. Entire launched its preview on July 8, 2026, with regional mirrors in the US, EU, and Australia. The pitch is simple: your code stays on GitHub while agents clone from a regional mirror that can handle high concurrency without hitting rate limits.

This guide covers setup, architecture, and the features that matter for teams running Claude Code, Codex, Cursor, or other agentic workflows at scale.

Why distributed Git for agents#

Centralized Git hosting works fine for human developers. A few clones and pushes per hour per engineer does not stress GitHub's infrastructure. AI coding agents are a different load profile. A single fleet of agents running parallel sessions can issue thousands of clone and pull operations per hour against the same repository.

Dohmke frames this as a fundamental constraint in an Entire blog post: the strain shows up as rate limits, high latency, or outages. Entire's answer is regional mirrors that absorb heavy read traffic while keeping the origin repository on GitHub as the source of truth.

How Entire works#

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.

The workflow is straightforward:

1. You install the Entire GitHub App and grant access to the repositories you want to mirror.
2. You create a mirror for each repository.
3. Agents clone from the regional Entire URL instead of GitHub.

The mirror stays in sync with GitHub. When agents need to push, they can write to the Entire mirror, and changes propagate back to origin. The practical difference is that your agents hit a regional endpoint optimized for high-volume read operations, not a centralized service serving the entire planet.

Setup#

Install the CLI:

# Homebrew
brew install --cask entire

# curl
curl -fsSL https://entire.io/install.sh | bash

Authenticate:

entire login

Create a mirror for an existing GitHub repository:

entire repo mirror create

The CLI walks through repository selection interactively. Once the mirror is created, clone from it:

entire repo clone /gh/OWNER/REPO

Alternatively, clone using a direct regional URL:

git clone entire://aws-us-east-2.entire.io/gh/OWNER/REPO

The entire:// protocol uses a custom gitremote-helper included with the CLI.

Regional architecture#

Entire currently operates in three regions: US (aws-us-east-2), EU, and Australia. Users can pin data to a single region or spread across multiple for redundancy.

The architecture prioritizes:

• Scale: Regional API nodes and distributed object storage handle concurrent operations.
• Latency: Read operations (refs, commits, files, diffs, merge bases) resolve against regional data.
• Availability: Multi-zone replication with automatic repair and catchup flows.

For teams with data sovereignty requirements, regional pinning keeps code within geographic boundaries while still enabling collaboration.

Performance#

Entire's benchmarks claim approximately 570,000 clones per hour from a single repository and around 586 pushes per second (2.1 million hourly). Those are throughput numbers under load, not latency figures, but they indicate the system is designed for the agent-era workload pattern.

The company released ForgeMark, an open-source MIT-licensed benchmarking tool, alongside the announcement. If you want to validate these claims against your own workload, ForgeMark is the tool to use.

Agent integrations#

Entire integrates with the major coding agents:

• Claude Code: Point your project at the Entire remote.
• Codex: Configure the Entire mirror URL in your workspace settings.
• Cursor: Use the Entire remote in your project's git config.
• Factory AI: Mirror your repositories to Entire and reference the regional URLs in your Factory configuration.
• GitHub Copilot: Continues to work normally since the code still lives on GitHub.

The integration is Git-level, not tool-specific. Any agent that clones via standard Git commands works with Entire mirrors.

Semantic memory layer#

Beyond raw performance, Entire captures metadata that GitHub does not: agent sessions, prompts, and tool calls. This data is stored alongside the code.

Three features use this context:

• Entire Blame: Surfaces the agent session, prompt, and decision behind a line of code.
• Entire Review: Sends branches to multiple agents in parallel for intent-aware code review.
• Semantic search: Query why code was written, not just what changed.

For teams debugging agent-generated code, this audit trail matters. When a bug surfaces six months later, you can trace back to the prompt and reasoning that produced the problematic line.

Branches and writes#

Mirrored branches stay in sync with GitHub. For high-volume write workflows, Entire supports "Entire-native branches" with higher concurrency limits.

If you need to write without syncing back to GitHub, use the entire/unmirrored/ branch prefix:

git push origin entire/unmirrored/experiment-branch

These branches stay regional and do not propagate to the GitHub mirror. Useful for scratch work, experiments, or intermediate agent outputs you do not want in the main repository.

What is coming#

According to the launch documentation, the roadmap includes:

• Native public and private repository hosting on Entire (not just mirrors)
• Open-source Git backend with self-hosting support
• Tamper-evident branch history and policy-as-code protection
• CI/CD pipelines
• Enterprise organization management and policies

The company is currently waitlisting new users to manage capacity as they scale.

When to use Entire#

Use Entire if:

• You run agent fleets that hammer the same repositories with concurrent clones and pulls.
• You hit GitHub rate limits during agentic workflows.
• You want an audit trail of agent sessions and prompts tied to code.
• You need regional data residency for agent operations.

Skip Entire if:

• Your agent usage is light and GitHub's limits are not a problem.
• You do not want another service in your stack.
• You need features that are still on the roadmap (native hosting, CI/CD).

FAQ#

Is Entire replacing GitHub?#

No. In the current preview, Entire mirrors GitHub repositories. Your code stays on GitHub as the source of truth. Entire handles the read-heavy agent traffic. Native hosting is on the roadmap but not yet available.

What does Entire cost?#

Pricing is not public yet. The preview is accessed via waitlist, and the company has not announced tiers or per-seat costs.

Does Entire work with private repositories?#

Yes. You authorize access via the Entire GitHub App, then create mirrors for private repositories the same way you would for public ones.

Can I self-host Entire?#

Not yet. The roadmap mentions open-sourcing the Git backend with self-hosting support, but that is not available in the preview.

How does Entire handle conflicts between agent pushes?#

Entire is Git-compatible, so standard Git merge and conflict resolution applies. Multiple agents pushing to the same branch will encounter merge conflicts the same way human developers would.

Does Entire change how I use Claude Code or Codex?#

No. You point your project at the Entire remote URL instead of GitHub. The agents clone and push using standard Git commands. No code changes required.

Official Sources#

Continue Reading#

• Agents 101: How to Build and Deploy Anything with AI Agents
• Cloudflare DDoS Report H1 2026: 1 Tbps Attacks Soared as DNS Floods Became the Leading Vector
• Flagship: Cloudflare Feature Flags for AI Apps
• Multica Turns Coding Agents Into Teammates. The Hard Part Is Receipts.
• LLMs Resolve Java Merge Conflicts Better Than Structured Tools - Because They Never Give Up