The short version
The company founded by GitHub’s former CEO is betting that the pull request does not survive AI coding agents. Entire’s new forge has agents score every push against team-defined checks, so people approve results instead of reading every line. If your CI/CD and delivery pipeline already struggles with AI-generated changes, the real lesson is to write down your review rules as automated evaluations, whichever platform you use.
What did Entire launch?
Entire describes the product as a Git forge: a place to host code, review changes and run builds, built for teams where agents write most of the code. It has three parts. Trails replaces the pull request. Distributed Git hosting is a Git-compatible server with private repositories in four regions. Continuous Memory syncs coding-agent sessions into the repository so later agents can search what earlier ones did and why.
Builds run through integrations with Buildkite and Depot rather than an in-house CI system. Entire has also open-sourced two pieces of the memory layer, a codebase graph and an “agentic brain” that turns past sessions and prompts into queryable memories. Self-hosting, public repositories and more partner integrations are listed as coming later. The launch announcement did not detail pricing.
How do Trails replace the pull request?
In a classic pull request, a reviewer reads the diff line by line. In Trails, a team defines evaluations, for example security rules, test coverage, API compatibility or style. On every push, review agents run those checks and post scores and findings to a dashboard. Coding agents address the findings in a loop, and the developer watches the scores move. The merge gate opens only when the evaluations are green, required humans have approved and CI passes.
The pitch is speed: Dohmke argues that agents now write almost all of the code, faster and in larger changes than people can read. The risk is equally clear. An evaluation only catches what someone thought to measure, and a dashboard of green gauges can hide a design mistake that a human reader would have flagged.
Why are AI coding tools rebuilding source control?
Volume is breaking old workflows. Anthropic said in October that merges to its main repository rose from about 2,500 a day in May to more than 30,000 a day by September, and it moved that monorepo from Git to an in-house system. Cursor launched its own hosting service, Origin, in August. Entire positions itself against both and against GitHub, citing an incident-monitoring analysis that counted 257 GitHub incidents in the 12 months to April 2026.
The timing is pointed: GitHub opens its Universe conference in San Francisco on October 28, with agent-driven development as its main theme. For buyers, competition means faster improvements in review tooling on every platform, not only on the newcomers.
What it means for US & EU software teams
The practical shift is from reviewing code to reviewing evidence. Teams that already use AI coding agents should expect review time, not typing time, to decide delivery speed. That puts value on things any platform can use: written acceptance criteria, automated security and dependency checks, contract tests for APIs and clear rules on which changes always need a senior human.
For regulated work, keep an audit trail. If a merge decision rests on agent scores, record which evaluations ran, which model produced them and who approved the result. EU teams should also check where agent sessions are stored: Continuous Memory keeps prompts and reasoning in the repository, which may contain customer data, so regional hosting and retention rules matter for GDPR. Entire’s EU region helps, but residency should be confirmed in the contract, not assumed.
How should teams evaluate an agent-first forge?
- Pilot on one service. Mirror a non-critical repository and keep your current host as the system of record.
- Write the evaluations first. Turn your review checklist into automated checks; they work on GitHub and GitLab too.
- Keep humans on high-risk paths. Require named approvers for auth, payments, data migrations and infrastructure code.
- Test the vendor claims. Measure push times, CI duration and agent token use on your own codebase.
- Check data handling. Confirm region, retention and access controls for stored agent sessions before enabling them.
- Plan the exit. Make sure repositories, history and review records can be exported in standard Git form.
Frequently asked questions
What did Entire launch on October 26, 2026?
Entire, the startup led by former GitHub CEO Thomas Dohmke, launched a Git forge, a service for hosting code, reviewing changes and running builds. It combines Trails, an agent-scored review flow that replaces the pull request, distributed Git hosting with private repositories in the US, EU, India and Australia, and Continuous Memory, which stores coding-agent sessions in the repository.
How do Trails differ from pull requests?
Instead of a person reading every changed line, review agents run custom evaluations on each push and return scores and findings. Coding agents fix the findings in a loop, and the merge gate waits for green scores, human approvals and passing CI. Humans still approve the merge; they review the scores and the risky parts rather than the full diff.
Are Entire’s speed and token-saving claims verified?
Not independently. Entire says its hosting pushes up to 89 times faster than GitHub and six times faster than Cursor Origin, and that its codebase graph cuts agent token use by up to 71%. These are company benchmarks. Teams should test with their own repositories, CI and agents before relying on them.
Should we move our repositories off GitHub now?
Not yet for most teams. Entire is days old, self-hosting and public repositories are still on its roadmap, and pricing was not detailed at launch. A low-risk path is to pilot it on one internal service while keeping GitHub as the system of record, and to define the evaluations and merge rules you would want on any platform.
Sources
Entire — Entire launches an independent Git forge for agent-human collaboration (primary source, October 26, 2026)
GeekWire — Former GitHub CEO launches rival platform built for AI coding agents (October 2026)
The New Stack — Ex-GitHub CEO launches a GitHub rival that reinvents the pull request
Tech Times — GitHub had 257 incidents in 12 months: Entire ships distributed forge for AI agents (October 27, 2026)
TechCrunch — Former GitHub CEO raises record $60M dev tool seed round at $300M valuation (February 10, 2026)
InfoQ — How Anthropic redesigned source control for AI-driven development (October 2026)