Former GitHub CEO Launches Git Platform Built for AI Agents

Entire, a startup founded by former GitHub CEO Thomas Dohmke, has launched a Git-based developer platform that uses AI agents to review code, manage repositories and retain context from previous coding sessions.

The platform, called Git Forge, is built for agent-human collaboration. Announced on October 26, the platform introduces three core features: Trails, an agent-based code review system; distributed Git hosting; and Continuous Memory, which stores coding sessions and prompts in repositories for AI agents to retrieve later. 

The platform is available to developers, with repository hosting in the US, Europe, Australia and India.

Entire claims its source control server can process code pushes up to 89 times faster than GitHub and six times faster than Cursor Origin, based on the company’s own benchmarks.

Entire replaces pull requests with Trails

Trails aims to make code reviews easier as AI coding agents generate more code than developers can review manually.

Traditional pull requests rely on code diffs, comments, and continuous integration (CI) results to help developers decide whether to merge changes. Entire argues that this workflow becomes harder to manage when coding agents generate changes and other agents repeatedly review them.

Trails brings evaluation scores and unresolved findings to the top of the review interface. Whenever code is pushed, review agents assess the changes independently in cloud sandboxes and return scores, explanations and suggested fixes.

The system includes evaluations for security, code quality and whether human review is required. Developers can configure additional checks and set rules, called Gates, that prevent merges until specified conditions are met, including human approvals and passing CI checks.

“The goal is simple: Give all the information you need to decide whether enough loops between your coding agent and the review agents have happened to achieve a stable state. Or not, and another iteration is needed. Only when the confidence is high, risk is low, and all findings are cleared, the merge should happen,” said Dohmke.

Entire uses Claude Code and Codex to run its review workflows, with isolated cloud environments provided by Daytona and E2B.

Distributed hosting targets AI coding workloads

The platform also allows developers to host repositories directly on Entire or mirror existing repositories from GitHub. Teams can choose a primary hosting region and replicate repositories across regions to meet data residency requirements.

Entire says its distributed architecture is built around a Git-compatible source control server, a distributed object store and a transactional reference database. Its internal benchmarks report 418 pushes per second, compared with 4.7 for GitHub.

Developers can continue using GitHub as their source of truth while running Trails on Entire. The company also provides a migration option for teams that want to move repositories to its platform.

For continuous integration, Entire has partnered with Buildkite and Depot. Both services can run build and test workflows when code is pushed, with results displayed in Trails. Failed builds can block merges, while coding agents can use the services’ interfaces to address problems.

Continuous Memory retains coding context

Entire’s third component, Continuous Memory, is intended to prevent coding agents from losing project context when a session ends.

The system comprises two open-source tools. Entire Graph maps a codebase, including definitions, callers, types and dependencies, to help agents find relevant information. Entire Brain links that map with previous agent sessions and prompts, allowing agents to retrieve the reasoning behind earlier code changes.

The company claims Entire Graph reduced token consumption by up to 71% in its tests.

Entire Brain can also connect to the company’s backend, allowing teams to share agent memories rather than keeping them on individual machines. This could help agents account for earlier decisions, understand existing code and avoid repeating previously resolved mistakes.

Entire raised $60 million in seed funding at a $300 million valuation in February 2026. The round was led by Felicis, with participation from Madrona, Microsoft’s M12, Basis Set and other investors. The company described it as the largest seed round in developer tools at the time.

Dohmke launched Entire after leaving GitHub, where he served as CEO. The startup is building infrastructure around existing coding agents rather than developing its own foundation models.