Entire

Entire was founded by former GitHub CEO Thomas Dohmke and launched in February 2026 with a record $60M seed round at a $300M valuation led by Felicis. Its thesis: as AI coding agents produce the majority of code, the software development lifecycle must be rebuilt around human-agent collaboration rather than human authorship. The first product, Checkpoints, is an open-source CLI (Go, MIT license) that hooks into Git and automatically captures AI agent sessions — full transcripts, prompts, token usage, and AI-vs-human line attribution — on a separate orphan branch without cluttering the main codebase. It supports Claude Code and Gemini CLI at launch. Behind it sits a planned three-layer architecture: a Git-compatible database for AI-produced code, a semantic reasoning layer tracking decision trails, and an AI-native collaboration interface.

The honest picture: only the first layer (Checkpoints) has shipped; the reasoning layer and AI-native interface are roadmap. The CLI repo drew 2,700+ stars in 9 days but also 400+ issues, showing real early-adopter bugs. Skeptics note Anthropic or Google could build session capture directly into their CLIs, and "the concept of reviewing code is dying" is a provocative claim that many engineering leaders reject.

Install the Checkpoints CLI from the entireio/cli GitHub repo, run the init hook in your repository, and use Claude Code or Gemini CLI normally — sessions are captured automatically on commit.