Ex-GitHub CEO Dohmke Launches Entire, a Coding Platform for Agents

Your daily brief on leadership and business

Entire, a developer platform from former GitHub CEO Thomas Dohmke, is now live and open to all developers. It is built around coding agents and positioned as a rival to GitHub.

The platform sits on Entire's decentralized Git network, announced in July. That network is meant to cope with the sharp rise in AI coding activity, which has caused a growing number of GitHub outages. Entire describes its forge as "a GitHub alternative for the speed that teams require to merge changes as fast as coding agents create them."

Trails instead of pull requests

The first change is that pull requests give way to what Entire calls Trails. A pull request asks a reviewer to read a diff. A Trail gives developers a heads-up display, which the company likens to an airline pilot's, with custom evaluations that agents score on every code push.

The data is available in Entire's web interface and through its open-source CLI. Coding agents can work through the findings in a loop while the human developer checks progress. Once every evaluation passes, the code can be merged without anyone reading it line by line. The agents in this process can be customized, and each Trail shows how they spend tokens.

"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," Dohmke said. He added that multiple agents generate the evaluations, so developers can merge with confidence without reviewing every line of the diff.

Speed and memory

Entire claims Git hosting up to 89 times faster than GitHub and six times faster than Cursor origin, which it credits to the decentralized network. Developers can create, migrate and host private repositories in several regions, starting with the US, India, Australia and the EU.

A second feature, Continuous Memory, keeps developer session logs next to the code in the repository. Agents can search them for answers. Entire says this helps them use up to 71% fewer tokens.

The memory layer combines a codebase graph with what Entire calls an "agentic brain," which uses earlier sessions and prompts as lessons so that agents do not repeat mistakes.

Source: ZDNET