When Nick Holzherr talks about legal inefficiency, it’s not theoretical. “I’ve spent over a million dollars on legal contracts,” he tells me from his home office in Birmingham, England. “Ninety percent of it was just someone editing a template(slowly), and charging thousands.”
That frustration turned into GitLaw, a startup redefining how legal work is done by treating contracts like code, version-controlled, machine-readable, and open-source.
The company’s mission is deceptively simple: make legal documents as programmable as software. Instead of static Word files that pass endlessly between lawyers, GitLaw turns them into structured, editable repositories that AI agents, and humans, can safely collaborate on.
From Kitchen Recipes to Legal Code
Holzherr isn’t new to automation. He previously founded Whisk, a food-tech platform acquired by Samsung, where he spent years building recommendation engines for recipes and groceries. The experience of selling that business, and watching lawyers move slower than his servers, left a mark.
“Every contract took weeks when it should’ve taken days,” he says. “And every bill reminded me that the problem wasn’t complexity, it was inefficiency.”
When AI tools like ChatGPT started entering the mainstream, Holzherr noticed founders using them for legal tasks. “Everyone was hacking together NDAs or contracts with ChatGPT,” he says. “But those systems hallucinate. They’ll literally read the first fifty lines of your contract, miss the rest, and tell you it’s fine, confidently.”
That, he adds, is why GitLaw exists. “We’re not replacing lawyers. We’re replacing bad automation.”
Git for Law: Version Control Meets Legal Reasoning
At the core of GitLaw is a bold idea: apply Git-style version control to legal documents.
In practice, every contract lives as a structured file that records who changed what, when, and why. Instead of messy Word redlines, each edit is tracked, reversible, and auditable. The result is legal collaboration that looks, and behaves, like software engineering.
“Code has solved this,” Holzherr says. “In GitHub, you can see every commit, branch, and pull request. Why shouldn’t legal work be the same?”
Under the hood, GitLaw’s infrastructure is powered by a multi-step agent system. Each AI agent handles a defined part of the process, drafting, filling metadata, checking jurisdiction, or reviewing clauses, using different models to maintain both speed and precision.
Users see every AI edit in tracked changes, making the system transparent and verifiable. “You always know what the agent did,” Holzherr says. “It’s like having an AI paralegal whose work you can audit.”
The Data Flywheel: Market Standards at Scale
For Holzherr, the real moat isn’t the interface, it’s the data layer.
Every vetted contract contributes to an evolving understanding of “market standard.” If enough businesses use GitLaw, the system can detect when a clause deviates from norms and alert the user.
“Imagine your contract comes back with a note saying: this clause isn’t market standard, you should push back,” Holzherr says. “That’s where the real intelligence lives.”
To accelerate this loop, GitLaw offers a free tier and $20/month pro plan, mirroring ChatGPT’s model. “We want as many documents as possible in the system,” Holzherr explains. “That’s how we learn what’s normal, and make law transparent.”
Trust: The Real Differentiator
Holzherr is candid about AI’s limitations. “All models hallucinate,” he admits. “But our job is to limit that risk by grounding generation in vetted templates, showing revisions, and constraining the scope of edits.”
GitLaw’s templates come from trusted legal organizations like Common Paper and NVCA. Over time, he hopes to decentralize this vetting through community verification and upvoting by accredited lawyers.
He also hints at future trust mechanisms: AI-generated documents with insurance and optional human-in-the-loop verification for $20-$30 per review.
“Law firms are insured. We can be too,” he says. “Imagine a legal document that’s not just generated by AI, it’s insured like a contract created by lawyers. That’s where this is heading.”
A Platform Built for Agents
Perhaps the most forward-looking part of GitLaw is its API-first vision.
Soon, other AI systems, not just humans, will be able to call GitLaw’s APIs to generate, fill, and sign contracts automatically. “Your sales CRM could trigger a contract when a deal closes,” Holzherr explains. “Or your hiring agent could draft an employment agreement the moment a candidate accepts.”
If Whisk was about organizing food, GitLaw is about organizing law. Both follow the same pattern, find the messy, unstructured corner of the world and give it an API.
Holzherr sees the market as inevitable. “The biggest legal AI tools today, Harvey and Legora, are built for elite law firms,” he says. “They make expensive lawyers more efficient. But the 90% of small businesses who don’t have a lawyer? That’s who we’re building for.”
In that sense, GitLaw isn’t just a product. It’s a protocol for the legal future, the connective tissue between humans, AI agents, and the documents that govern them.
