Entire Review: Git-Compatible Platform for AI Agent Code Hosting with Trails

Entire is a Git-compatible code hosting platform built specifically for AI coding agents alongside human developers. Founded by former GitHub CEO Thomas Dohmke, it launched publicly on October 26, 2026 after a February 2026 company formation and a July 2026 preview of its distributed network. The service targets high-volume agent workflows that overwhelm traditional forges through distributed mirrors, agent-driven review gates, and persistent project memory.

Entire Code Hosting, Trails Review System, and Continuous Memory

Entire operates a distributed Git network with mirrors in Ohio, Frankfurt, Sydney, Mumbai, and additional locations under expansion. Repositories can be hosted natively or mirrored from GitHub while preserving the upstream as source of truth. Native branches remain functional during upstream outages. Benchmarks using the open-source ForgeMark tool show push rates reaching 418 per second, compared with roughly 4.7 on GitHub.

The Trails system replaces traditional pull requests. Each push triggers independent review agents that run configurable evaluations such as security risk, intent drift, low-quality output, and test coverage inside cloud sandboxes. Scores accumulate with run history, and merge gates enforce passing evaluations, CI checks, and human approvals. Trails support multi-repository grouping and integrate with CI providers including Buildkite and Depot.

Continuous Memory captures agent sessions including prompts, reasoning steps, tool calls, and decisions, linked to commits after local secret redaction. The open-source entire-graph component builds a semantic and symbol graph of code for retrieval, with reported token reductions up to 71 percent on LoCoMo benchmarks. entire-brain indexes sessions, graphs, and documentation into shared memory accessible via Agentic Search. An MIT-licensed CLI records sessions from supported agents through the Agent Integration Protocol.

Entire Pricing, Ease of Use, Security, and Comparisons

Core platform hosting remains free during the launch period. Charges apply only for review and agent model token usage at the rates of the underlying providers such as Anthropic and OpenAI. Each billing account receives a one-time $100 starter credit valid for 90 days or until exhausted. No per-user or storage fees are published for core repository features. Enterprise plans require direct sales contact.

Registration occurs through the web interface at entire.com. The service supports standard Git workflows plus the open-source CLI for session capture. Remote team size stood at roughly 40 employees in mid-2026 with plans to reach 60 by year end. Data residency options align with the listed regional mirrors. No major security incidents have been reported since launch, and the platform relies on provider-level model security plus local secret redaction.

Compared with GitHub, Entire adds native agent review agents and persistent memory but lacks GitHub’s mature ecosystem of third-party apps. GitLab offers self-hosted options and built-in CI yet does not include agent-scored Trails or session graphs. Sourcegraph focuses on code search without distributed hosting or review gates. Entire’s 89× push throughput advantage appears only in agent-heavy workloads.

Limitations include the absence of published paid tiers beyond token usage, reliance on third-party model providers for reviews, and preview-stage status of some CI integrations. Production hardening of merge gates remains ongoing according to documentation. User reports on public forums note faster clone times for agent fleets but occasional delays when review agents queue during peak hours.