The idea

Git solved the problem in software

Developers solved collaboration two decades ago. Not with meetings, but with Git . Every change is tracked, diffed, signed, reversible, and blameable.

The way I work, I don’t trust everybody. In fact, I am a very cynical and untrusting person. The whole point of being distributed is I don’t have to trust you.

— Linus Torvalds, Creator of Git

Outside software

Outside of software, it is a very different story. Versions are stored inside a single domain (SaaS) or per file document, spreadsheets and design files. This leads to an isolated history, no shared timeline, no unified view of how a project evolved.

Versioning isn’t the foundation of collaboration and workflows — it’s an afterthought. While developers work in branches, everyone else works in final_v3_really_final.docx.

Conflicts and interference are “solved” by communication: Endless meetings, human-to-human coordination within a shared layer of trust. Nobody likes the current environment with systems like SharePoint, but the setup is just good enough to resist real change.

Agents make the problem big

Now enter agents — working in parallel, lightning-fast, but unreliable and fundamentally untrusted. The old human-centric setup can’t keep up. It’s already starting to crack.

Projects with thousands of untrusted collaborators already exist — and they run the backbone of modern software. Linux, Postgres, and countless others thrive under this exact model. They’re all managed with a system built for that kind of chaos — Git.

What is missing? - L’ingrédient manquant à git

Git nails accountability and safety but it can’t enforce boundaries nor can track all information needed for coordination. Git detects changes after the fact - and it’s completely blind to read.

To enforce boundaries, a system must intercept actions, not just watch them. For coordination, it must know what was read, in which state, and what assumptions were used to produce an output. This information is essential to detect conflicts before they happen.

How Legit Fills the Gap?

Legit is designed to be the infrastructure for AI collaboration. It doesn’t watch — it intercepts. Reads and writes can be tracked, intercepted, and even rejected when they violate boundaries or assumptions.

All of this is anchored in a solid, Git-backed data layer, ensuring that every action — whether accepted or blocked — is auditable, versioned, and reversible. Legit brings control and coordination to AI agents while standing on the strong foundation that made Git the backbone of software collaboration.

Story Beat

“When I (Nils) worked at Decipad , we let users co-create financial reports with AI. At first, it was magical. AI could rewrite, analyze, and summarize on command.

But in high-stakes reports, the magic broke fast. A single AI edit could rewrite assumptions or distort results, with no way to trace or undo it.

That’s when it clicked: If AI can change the source of truth, it must also be version controlled by design. Version control isn’t optional — it’s survival.”

Takeaway

AI moves fast and brings new requirements to the table, where trust can’t be an afterthought. Legit brings Git’s discipline (versioning, traceability, and accountability) to the new world of AI-driven collaboration.