- Thomas Dohmke launched Entire in February 2026 with $60 million in seed funding and a $300 million valuation to rebuild version control for AI agent workflows.
- Entire’s open-source Checkpoints CLI extends Git to store prompts, session transcripts, and tool calls as structured metadata for AI-generated code.
- Entire’s distributed Git network sustained 331 pushes per second in its own testing, compared to 8 per second on GitHub, a self-reported benchmark.
Thomas Dohmke spent years running GitHub, the infrastructure underpinning how most developers collaborate. Now he’s betting that infrastructure needs to be rebuilt from scratch for a world where AI agents are doing the coding. His new company, Entire came out in February 2026 with $60 million in seed funding and a $300 million valuation.
His new company, Entire, has launched with seed funding and a reported valuation.
Rethinking Git for AI Code
The core problem Dohmke is solving is straightforward: Git was built for humans. When dozens or hundreds of AI agents are concurrently cloning, pulling and pushing code, centralised Git hosts hit rate limits fast. Worse, traditional version control has no way to capture why an agent made a change, just that it did. Human reviewers are left trying to reverse-engineer intent from a diff, which is slow and error-prone.
Entire’s answer is an open-source CLI tool called Checkpoints that extends Git’s capabilities. It stores not just code but the full interaction history behind it: prompts, session transcripts, modified files, token usage and tool calls, all as structured metadata tied directly to commit SHAs. Every AI-generated contribution becomes traceable and reviewable, with a clear record of the reasoning behind it.
Distributed Network and Agent Context
On top of raw throughput, Entire adds a semantic memory layer that logs agent sessions, prompts and tool calls directly into the repository. Two features built on this are worth noting for teams building agentic pipelines: “Entire Blame” links any line of code back to the specific agent session and prompt that produced it, and “Entire Review” handles multi-agent diff analysis on pull requests. There’s also semantic search across code history by intent rather than text, useful when you need to find not the line that changed, but the reason it did. If you’re building with tools like LangChain or CrewAI and hitting the review bottleneck this kind of audit trail is designed to address, that’s where Entire is aiming.
Enterprise Integration and Future Outlook
The platform has three components: a Git-compatible database built from scratch, a semantic reasoning layer for multi-agent collaboration, and an AI-native UI for agent-to-human workflows. It currently integrates with GitHub Copilot, Claude Code, Codex, Cursor, Factory AI and Gemini, covering most of the coding agents teams are already running.
Pricing will follow the preview period, with seat-based and consumption-based tiers, a free tier and open-source components. Backers include Felicis Ventures and Microsoft’s M12. Whether Entire becomes standard infrastructure for agentic development or a niche tool for teams already hitting Git’s limits is the real question, but for anyone managing serious AI agent workflows today, the audit trail problem it’s solving is real. For more on AI agents and automation tools, visit our AI Agents section.



