Entire launches agent-first Git platform with automated evals

By Rebecca Stone

Entire is live today, and the developer platform claims Git hosting “up to 89x faster than GitHub” and six times faster than Cursor origin.

Entire is a GitHub rival built around coding agents. The platform comes from former GitHub CEO Thomas Dohmke. ZDNET reported that AI has sped up coding, placed new demands on infrastructure and changed how developers work. The platform’s stated purpose is to remove coding bottlenecks created as AI-written code changes what developers spend time on.

The stakes are operational, not merely editorial. Entire says AI-driven coding has increased pressure on code-hosting infrastructure and contributed to mounting GitHub outages. Entire announced its decentralized Git network in July, built the platform on it and designed the network to withstand that surge.

The launch makes infrastructure and review design one product decision. Entire places agents inside the path from a code push to a merge rather than treating them as optional assistants outside the workflow. That is the commercial test for a platform serving developers whose output now includes AI-generated code.

What does Entire replace with Trails?

The first change Entire makes to the standard coding workflow replaces pull requests with a feature the company calls Trails.

Entire said: “Unlike the classic Pull Request model of reviewing code changes with a diff, Trails offer developers a heads-up display similar to an airline pilot, with customized evals that are scored by agents on every code push.”

Entire presents the proposed idea, the agent trace and the session outcome as first-class artifacts for review alongside the code. The design shifts the default task from inspecting a line-by-line diff to validating agent-generated evaluations before a merge. The quality of those evaluations now determines whether a human can approve code without reading every generated line.

Entire treats review as a feedback system. A push produces evaluations, agents address findings, the developer validates progress, and the merge follows once every evaluation passes.

Entire said: “All data is available not only on Entire’s web interface, but also through the open-source Entire CLI. This allows coding agents to address findings in a loop, while the human developer validates the progress. When all evals are satisfied, they can merge the code without needing to look at the code itself.”

The open-source Entire CLI gives agents access to findings they can address in a loop, while Entire’s web interface exposes the same data to developers. Under the company’s workflow, an agent works through failing findings, the developer validates the progress, and the code can merge once every evaluation passes. The model reduces manual diff review, but it also makes the agent’s evaluation process part of the merge gate.

“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.

“It uses multiple agents to generate evals that verify the quality and success of AI-generated code to let developers confidently merge without reviewing every line in the diff.”

Entire divides the record into the idea that started the work, the trace of the agent run and the outcome of the session. Multiple agents generate evaluations to verify both the quality and success of AI-generated code. That makes the evaluation system, rather than the raw diff, the proposed control point for review.

The agents running the process are customizable, and each Trail shows how agents spend tokens. Entire therefore places both the outcome and token consumption of automated work inside the review record. That telemetry is specific to each Trail and tied directly to the code-review sequence.

How does decentralized Git change hosting?

Entire says developers can create, migrate and host private repositories in multiple locations around the world. The initial geographies are the US, India, Australia and the EU.

The launch gives private-repository customers four named starting geographies. It also makes migration an explicit product operation rather than an implied consequence of joining. The report did not provide a schedule for adding locations.

Entire attributes its hosting claims—“up to 89x faster than GitHub” and six times faster than Cursor origin—to the decentralized network. For a GitHub rival, those figures can support a migration pitch only if customers reproduce them with their own repositories and agent activity.

ZDNET’s report included no benchmark methodology or independent validation for either comparison. It did not define the workloads, repository sizes or measurement process behind the figures. The speed claims therefore establish a target for prospective users rather than an independently verified advantage.

Multi-location hosting also gives organizations a concrete deployment choice. Entire says developers can move private repositories among the available locations and host them there. The report did not explain how repository data moves between those locations or what operating constraints apply during a migration.

What does Continuous Memory remember?

Continuous Memory keeps developer session logs next to code in the repository and uses them as a data source. Coding agents can search those records for answers, and Entire says the approach helps them use up to 71% fewer tokens.

ZDNET compared the feature with memory tools in AI chatbots. Those systems reuse past conversations or other agent interactions to reduce the time spent gathering context and speed query responses.

Entire identifies two components in Continuous Memory: a codebase graph and an “agentic brain.” The company says the brain uses previous sessions and prompts as lessons so agents do not repeat mistakes.

The codebase graph supplies the repository structure, while the agentic brain adds the history of earlier sessions and prompts. Together, they give an agent information beyond its current chat window and retain earlier corrections as part of the repository record.

The key distinction is location. Chatbot memory keeps context in an assistant interaction, while Continuous Memory anchors session logs beside the code they concern. Developers can review that history alongside the project, and agents can reuse previous work instead of beginning each task without context.

The token claim carries weight because repeated agent loops determine whether large-scale code review remains practical. Entire did not identify the baseline for the 71% figure, the workloads tested or a breakdown of the result. The reported savings remain a company claim without supporting test data.

What remains open at launch?

Entire is available to try today for all developers, according to ZDNET. The launch combines Trails, decentralized hosting and Continuous Memory in a platform for developers working with AI-generated code.

ZDNET’s report did not list pricing, paid tiers, capacity limits or independent results for the speed and token-saving claims. Those omissions leave the cost of adoption and the reliability of the benchmarks unresolved in the available material.

The product is accessible without a stated fee in the report, but price is not the only adoption variable. Teams must also decide whether agent-generated evaluations provide enough assurance to replace manual inspection of every diff. Entire’s answer is to give developers the trace, token use and outcome alongside the code.

For coding teams, the design could replace the familiar pull-request bottleneck with an automated quality gate, but only if those evaluations remain reliable. Entire has opened the platform to all developers; the next proof point is whether outside teams reproduce its speed and token savings while relying on agents to validate AI-written code.