Former GitHub CEO takes on GitHub: Entire launches a code platform built for AI agents
Thomas Dohmke's company Entire replaces pull requests with agent-scored Trails, lets customers keep code in the EU and claims pushes are up to 89 times faster than on GitHub. Pricing is not disclosed.
Entire, the company led by former GitHub CEO Thomas Dohmke, has launched its own code platform built for humans and AI agents working in the same repositories. The platform is open to any developer from Monday 26 October, and the company is taking direct aim at GitHub: Entire claims that pushes are up to 89 times faster than on GitHub, based on its own benchmarks.
Entire announced the launch in a blog post signed by Dohmke and on the company's X account, where the post went out at 18:11 Oslo time on Monday.
Pull requests are replaced by "Trails"
The core of Entire is what the company calls a Trail. It replaces the classic pull request, where changes are reviewed as a long timeline of comments with a CI status at the bottom. Dohmke writes that the pull request has become "an audit log that only the agent reads", because coding agents and review agents fill the timeline with long descriptions.
A Trail puts scores at the top instead. Each time new code is pushed, Entire starts review agents that run independently in isolated cloud sandboxes and return a score and a rationale. The company itself uses evaluations for security, "slop" (low-quality code) and whether human review is required. According to Entire, the agents use Claude Code and Codex with the latest models, in sandboxes provided by Daytona and E2B.
Customers can set up "Gates" per repository: requirements that must be met before code is merged. These can include human approval, passing CI checks or all agent findings being closed. Entire is partnering with Buildkite and Depot for continuous integration, and both offer tooling to migrate existing GitHub Actions workflows.
Faster hosting and code in your own region
Entire runs on a new Git-compatible source control server with distributed object storage. The company's own published benchmarks show 418 pushes per second against 4.7 for GitHub. The figures have not been independently verified.
For European organisations, the choice of region is the most concrete part. Customers can keep a repository in one region or replicate it to several. Entire says the service is live in the US, the EU, Australia and India, and cites data residency requirements as one reason to pick a region.
The company does not require customers to leave GitHub. A repository can be mirrored from GitHub so that GitHub remains the source of truth, and Trails can run on top of ordinary pull requests. According to Entire, the mirror keeps serving code if GitHub is down. Full migration is also possible, but the one-click option to detach a mirror from GitHub applies only to mirrors in the United States, according to the announcement.
Agent memory is stored in the repository
The third part is what Entire calls "Continuous Memory". The idea is that knowledge from agent sessions, such as decisions, attempts and why a line of code changed, is stored alongside the code instead of disappearing when the session ends. Two open-source tools are meant to do this: entire-graph, which builds a map of the codebase, and entire-brain, which links past agent sessions to that map. Entire says entire-graph saved up to 71 per cent of tokens in the company's own testing.
The announcement says nothing about pricing, and it does not give customer numbers or security and privacy certifications.
What it means for leaders
For organisations that already let AI agents write code, the launch points to a real problem: the review process was designed for people, not for a high volume of agent changes. Entire's answer is automated scoring and fixed requirements before code is merged.
It is still a new and largely untested vendor. Before moving source code, IT leadership should clarify pricing, the data processing agreement, where agent sessions are stored and who can access them. Mirroring from GitHub makes it possible to test Trails and the EU region without switching platforms. The speed figures should be measured on your own repositories before they are used to make decisions.
Sources and media
Entire's blog post is dated 26 October 2026. The company's post on X went out at 18:11 Oslo time the same day. All details and quotes in this article come from these sources, apart from the fact that Dohmke previously led GitHub.
Illustration: AI-generated with Grok Bot / hogby.ai