Thomas Dohmke, GitHub's former chief executive, launches Entire, a code platform built for working with AI agents
IA en un minuto newsroom · Editor: Jon Elgezabal
In 30 seconds
GitHub's former boss, Thomas Dohmke, is opening Entire to every programmer: a service for storing and reviewing software where AI agents and people work together. With Trails, several agents score each incoming change and someone on the team decides whether it goes in. It keeps agent sessions in the repository itself, can copy projects from GitHub and offers four regions, India and Australia among them.
Entire, Thomas Dohmke's company, launched its full platform on Monday, October 26.
Dohmke, co-founder and chief executive of Entire, previously ran GitHub, as noted by RuntimeWire. What he is presenting is a "Git forge": a service for hosting and reviewing code, in the mould of GitHub or GitLab, built so that AI agents and people can work together. The company says any developer can already use it.
Classic review compares changes line by line. Entire replaces it with Trails: whenever new code arrives, several review agents analyse it separately in isolated cloud environments, on criteria such as quality or security, and their scores appear at the top of the page so the person can decide. Each team can set requirements before changes are merged, such as human approvals or automated tests passing. For that review, Entire relies on Claude Code and Codex. Continuous integration testing comes through two partners, Buildkite and Depot.
The platform adds Continuous Memory, which keeps agent sessions and the instructions they received inside the repository, so other agents can look them up later. It is backed by two open-source add-ons that run locally. One of them, entire-graph, draws a map of how the code fits together, and in the company's own tests it cut token use by up to 71%.
Repositories can be created on Entire or mirrored from GitHub, and they are hosted in regions in the United States, the European Union, Australia and India. With a mirrored repository, if GitHub goes down, agents can still download the code from the regional copy. The company states, with tests it has published openly, that code pushes run up to 89 times faster than on GitHub and six times faster than on Cursor Origin. These are Entire's own measurements. The announcement does not mention pricing.
Why it matters · analysis and opinion
The bottleneck in agent-written software is no longer writing it but reviewing it, and Entire is betting that the classic change review cannot keep up with that volume. That is why it lets other agents do the scoring and leaves the final word to a person. The move carries the weight of someone who ran GitHub, but it enters a field where GitHub and GitLab hold almost every team. Its way in is mirroring: it can be tried without moving anything. Before switching homes, it is worth waiting for independent measurements of speed and savings, learning what it will cost and checking whether the agents' scores help people decide or just add another layer of noise.
Official source: Entire · Written with the help of AI: how we make the news
