The Pulse
Thomas Dohmke's Entire Launches a Git Forge for AI Agents
Entire, founded by former GitHub CEO Thomas Dohmke, launched a Git-compatible code forge for developers and AI agents on October 26, 2026. Its Trails review system, distributed hosting and session memory are designed to help teams manage ag
AI.info Team · Written with AI · fact-checked ·
“With Trails, Entire treats the idea, the agent trace, and the outcome of a developer’s sessions as first-class artifacts for review alongside the code.”
Thomas Dohmke, CEO of Entire
Entire launched its Git forge for agent-human collaboration on October 26, presenting it as a new home for code repositories and a review workflow built around AI coding agents. The company was founded by former GitHub CEO Thomas Dohmke, who now leads Entire. Its central pitch is that teams need to inspect not just code changes, but also the intent and agent activity behind them.
The launch combines three products: Trails, a review system that scores changes; Git-compatible distributed hosting; and Continuous Memory, which stores agent sessions and prompts alongside repositories. Entire says the service is available to any developer, with features aimed particularly at teams that use coding agents.
Trails scores changes on every push
Trails is Entire’s alternative to the conventional pull request. On each code push, review agents run evaluations and surface scores, findings and suggested fixes. Teams can enable built-in checks for code quality and security, as well as configure their own evaluations.
Developers can send agents back to address findings, while human reviewers see unresolved issues and can comment or dismiss them. Repository-level “Gates” let teams set conditions for approval, including human sign-off and passing CI checks. The system is intended to keep review criteria visible as code changes, rather than burying them in a long sequence of comments.
Entire says Trails can cover work across multiple repositories and begin with a feature request or bug report, or directly from an agent. The company’s launch materials describe a workflow in which agents iterate on code until configured checks pass; they do not establish that passing automated evaluations alone guarantees a change is safe or correct.
Entire claims 418 pushes per second
The forge runs on a new source-control server that Entire says remains compatible with Git. The company’s published benchmark lists 418 pushes per second for Entire, compared with 65 for Cursor, 15 for Pierre, 10 for GitLab and 4.7 for GitHub. Entire describes its results as up to 89 times faster than GitHub and six times faster than Cursor Origin; those comparisons come from the company’s own open-source benchmark.
Developers can host repositories directly on Entire, mirror projects from GitHub or migrate them. Hosting and replication are available in the United States, the European Union, Australia and India, according to the launch announcement. A GitHub mirror can continue serving clones and fetches if GitHub has an outage, while agents push to separate Entire branches until service returns.
That arrangement gives teams a way to try the forge without immediately making it their sole source of truth. Entire says mirrored pushes still route through GitHub, so its independent hosting and throughput claims apply most directly to repositories hosted natively on Entire.
Session history becomes searchable context
Continuous Memory is designed to preserve information that might otherwise disappear when an agent session ends. Entire’s CLI records sessions and prompts, while two open-source components provide code mapping and a memory layer: entire-graph indexes repository structure, and entire-brain links that map with past sessions. The company says its tests show entire-graph can reduce token use by up to 71%.
Entire also launched integrations with Buildkite and Depot for continuous integration. Teams can configure builds to run on pushes and display their results in Trails; the company says existing GitHub Actions workflows can be converted for use with either partner. Dohmke’s broader bet is that review needs to account for how an agent reached a change, not just the final diff.
The launch is a product announcement, not evidence that teams can safely merge code without human review. Entire’s own workflow includes human approvals among the gates teams can require. Public repositories, self-hosted Git nodes and live-streamed sessions remain on the company’s list of planned features, while the launch offering centers on private repositories, agent review and regional hosting.