Key Takeaways
- Entire, founded by former GitHub CEO Thomas Dohmke, launched a distributed Git network preview in July 2026 to solve the rate-limiting and scaling failures that occur when multiple AI agents hit centralised code repositories simultaneously.
- The platform raised a $60 million seed round in February 2026 at a $300 million valuation, led by Felicis with participation from Microsoft’s M12, positioning Entire as infrastructure for AI-driven development rather than a GitHub replacement.
- Entire’s open-source Checkpoints feature pairs every AI-generated commit with the prompts, session transcripts and reasoning logs that produced it, giving teams a direct audit trail into agent intent rather than just code output. When dozens of AI coding agents clone, pull and push to the same repository at once, centralised Git infrastructure breaks. Entire the new platform from former GitHub CEO Thomas Dohmke, launched a distributed Git network preview this month to address exactly that problem, mirroring repositories across regional nodes so agents read from local copies and stop hammering a single origin server.
Git’s AI Agent Bottleneck
GitHub and its peers were built for human collaboration patterns: bursts of coding, review cycles, occasional merges. That rhythm rarely generates the sustained, high-frequency read and write operations that fleets of AI agents produce. Rate limits hit fast, performance degrades and the whole pipeline stalls.
The core Git architecture was never designed for autonomous, always-on agents generating a continuous stream of commits. Developers caught in this mismatch either throttle their agents or accept an unstable environment. Neither squares with the accelerated development AI agents are supposed to deliver. The volume of automated pull requests also overwhelms review processes, creating what Dohmke describes as a “review bottleneck.”
How the Distributed Network Works
Entire mirrors GitHub repositories across regional nodes in the US, EU and Australia. Agents read from the nearest local copy, cutting latency and eliminating the congestion that hits when everything routes through a single origin. Developers connect their existing GitHub repos to Entire, so agents get the optimised network without anyone abandoning their primary version control setup. The distributed architecture also lays groundwork for data residency controls, a real concern for enterprise deployments in regulated industries.
Resilience is as important as speed here. Heavy agent load shouldn’t halt automated development workflows, and Entire’s distributed approach is built to keep operations running under sustained pressure.
Tracing AI Intent: Checkpoints and Blame
Distributed infrastructure handles throughput. Traceability is the harder problem. Entire’s Checkpoints feature, released as open source, automatically pairs every AI-generated commit with the prompts, session transcripts and reasoning logs that produced it. Without something like this, understanding why an agent made a specific coding decision often means re-running entire sessions from scratch.
Entire Blame links any line of code back to the specific agent session and prompt that generated it. For codebases where human and AI contributions are intertwined, that granularity matters for debugging and auditing. Entire Review goes further, running multi-agent diff analysis on pull requests so other agents or human reviewers can collaboratively validate changes. Together, these tools shift the review process from inspecting raw code to interrogating agent intent, which is where the real work is when agents are generating most of the commits. Teams building with Entire’s agent platform get all three out of the box.
The Semantic Reasoning Layer
Dohmke’s broader argument is that file-and-folder Git interaction is becoming the wrong abstraction as agents generate code at higher levels. Entire is building three layers: a Git-compatible database, a semantic reasoning layer and an AI-native interface on top.
The semantic layer lets multiple agents collaborate over a shared codebase without collision, sharing intent and specifications rather than raw code. The interface includes semantic search, query by intent (“show me all code related to user authentication logic”) rather than by filename or keyword. As agent-contributed codebases grow, that kind of search stops being optional. Initial integrations target Claude Code, Cursor, Codex and Copilot, connecting into their memory layers and extending their capabilities inside a managed version control environment.
Funding and Market Position
Entire raised $60 million in February 2026 at a $300 million valuation. Felicis led the round, with Madrona, Basis Set and Microsoft’s M12 also participating. Felicis described it as the largest seed investment ever for a developer tools startup, according to the company.
M12’s involvement says something about the competitive dynamic. Entire positions itself as a layer above GitHub, not a replacement, focused on agent reasoning and human-agent collaboration rather than core version control. That framing makes Entire complementary to GitHub rather than a direct threat, which likely made the Microsoft arm’s participation easier to structure.
Open Source and the Road Ahead
Entire plans to open source both its distributed Git backend and its benchmark suite. The initial release includes the Entire CLI, built to be portable and model-agnostic. Dohmke has said the platform will be stronger with community contributions, a philosophy consistent with his GitHub background.
The longer roadmap includes full decentralisation, native repositories that don’t require GitHub underneath, and advanced data sovereignty controls. Enterprise customers in regulated industries will care most about that last item. Whether Entire reaches full platform independence is a question of execution, but the $60 million seed gives the team runway to find out. For builders already hitting rate limits with agent fleets, the distributed network preview is available now.
Where Adoption Gets Hard
Developer inertia is real. Teams running modest agent workloads may not feel the pain Entire solves, at least not yet. Convincing them to add an infrastructure layer requires clear ROI and a frictionless onboarding experience, neither of which is guaranteed at preview stage.
Compatibility is the other pressure point. Integrations with Claude Code and Copilot are in place, but the agent framework space moves fast. Maintaining compatibility across LangChain, CrewAI, AutoGen and whatever ships next will take continuous effort. The semantic reasoning layer is early; its effectiveness depends heavily on the maturity of the models it connects, and interoperability standards for agent collaboration don’t yet exist. Pricing clarity will matter too, especially for enterprise buyers weighing new development infrastructure against incremental improvements to what they already run. More on where agentic AI is shifting infrastructure choices in our AI Agents section.
Originally published at https://autonainews.com/entire-launches-distributed-git-for-ai-agents-secures-60m-seed/
Top comments (0)