I am Roderick Lucas, founder and CEO of Moonlit. Before Moonlit I spent close to a decade at Deloitte, where I became Artificial Intelligence Lead, and I taught technology courses at Vrije Universiteit Amsterdam and Tilburg University. Moonlit is a legal publisher built for humans and AI. Legal professionals use our platform to research the law, read commentary and monitor changes. Underneath sits our Data Layer: legislation and case law from dozens of jurisdictions, connected by hundreds of millions of references, updated daily and verifiable down to the official source. The same layer reaches AI assistants such as Copilot, ChatGPT and Claude through our MCP connector, and other legal AI products through our API.<\/p>\n\n\n\n
The inspiration was a mismatch I saw every day. The law is public. You pay for it with your taxes, and the courts publish their rulings for everyone. Yet working with it was expensive and slow. Every serious question crosses a border at some point, the sources sit on dozens of national portals in different formats and languages, and the knowledge that explains them sits behind publisher paywalls. Specialists spent most of their time collecting and checking rather than thinking.<\/p>\n\n\n\n
Then language models arrived, and they could reason about legal text far better than anyone expected. What they lacked was something reliable to reason over. That gap was the company: build the layer underneath to provide access to the knowledge on top of it.<\/p>\n\n\n\n
Deloitte advises clients across many countries, and a tax or legal question rarely stays inside one jurisdiction. Our specialists were doing cross-border research by hand: open each national portal, search it in its own language, copy the sources into a memo. We built a tool to take that collecting work off their plates, with every source linked back to the official publication so the specialist could check it in one click.<\/p>\n\n\n\n
Once legislation and case law from several countries sat in one structure, and the citations between them were resolved into links, the tool stopped being a faster search box and became a map of the law.<\/p>\n\n\n\n
A data layer like this gets more valuable with every person and every system that reads from it, and inside one firm it could only ever serve one firm’s specialists. Together with Deloitte we decided spinning it out was the best path forward. We\u2019re in good contact and Deloitte is still an investor in Moonlit.<\/p>\n\n\n\n
The confidence came from the users, not from a plan. When people change how they work without being told to, you have something. Add to that the timing: language models were about to make verified legal data far more valuable, and we already had the building stones for layer those models needed.<\/p>\n\n\n\n
The first challenge was leaving a large organisation with a small team and a large ambition. Inside Deloitte you have brand, clients and infrastructure on day one. Outside you have to earn each of them again, and the legal market is careful about who it trusts. We answered that with proof rather than promises: ISO 27001 certification, EU and US hosting, and a product where every claim links to its source.<\/p>\n\n\n\n
We started as a research platform, became a data company when legal tech vendors began building on our API and MCP connector, and are now also a publisher with Commentaries. Each step was the right one, and each took some convincing, inside the company as much as outside it.<\/p>\n\n\n\n
The problems Moonlit solves fall into three different silos:<\/p>\n\n\n\n
Collection: The sources for a cross-border question are spread over dozens of national portals, each with its own format, language and way of citing. Moonlit connects to the official portals, takes in new publications every day and puts everything in one structure that is updated daily, down to the individual article.<\/p>\n\n\n\n
Connection: Legal texts cite each other constantly, and the meaning of a ruling depends on what it applies and what later rulings did to it. Our knowledge graph resolves those citations into links in both directions, so for any document you see what it cites and what cites it.<\/p>\n\n\n\n
Trust: Every document keeps its link to the official publication, and every answer our tools give can be checked in one click. A citation either points to a document we hold or it does not appear.<\/p>\n\n\n\n
Law firms, Big Four advisories and public institutions use the platform for cross-border research. Legal tech vendors build on our API and MCP connector instead of running a scraping operation per country. With Commentaries, anyone with a free account can read expert-reviewed commentary on the law.<\/p>\n\n\n\n
Against the traditional platforms, the difference is where we start. They were built per country and sold per seat, with the content licensed and locked. We start from the official sources of many jurisdictions, structured the same way and linked to each other, and we keep that layer open: through our platform for people, and through our API and MCP connector for machines. A legal tech company can build on it. A lawyer can connect it to the AI assistant they already use.<\/p>\n\n\n\n
Against other AI legal tools, we are not really a competitor. We are what many of them run on. Building and maintaining a complete layer per country is a scraping operation that never ends, and it is not what a legal AI company wants to spend its engineers on. So they take ours. Through the API and the MCP connector, any product or assistant can search our Data Layer. Our own assistant Luna runs on the same connector, so it gets no better data than theirs.<\/p>\n\n\n\n
And against everyone, the difference is Commentaries. We are the only one that we know of that publishes the interpretation of the law, reviewed by named experts, with every source one click away, free to read.<\/p>\n\n\n\n
Reliability starts at the source and never leaves it. We take the law from the official publishers only: the legislative gazettes, the court databases, the regulators and, for EU law, EUR-Lex. Every document keeps its attribution and a link back to the official publication, and that link is the basis for everything built on top.<\/p>\n\n\n\n
The concern is justified. Damien Charlotin’s tracker of court decisions involving AI hallucinations stood at well over two thousand when we launched Commentaries, hundreds of them tied to practising lawyers. A model has seen a case number mentioned far more often than it has seen the ruling itself, so it produces a reference that looks right. And its knowledge stops at its training date, while the law does not.<\/p>\n\n\n\n
Our answer is to change the job the model does. With our tools it retrieves the law instead of recalling it. Every result carries its citation and official link, and a citation that does not resolve to a document we hold does not appear. That removes the fabricated case. It does not remove misreading, which is why Commentaries puts a named legal expert between the AI draft and the reader. If you cannot click through to the ruling a conclusion rests on, you are being asked to take it on faith. We built Moonlit so nobody has to.<\/p>\n\n\n\n
We did not convince them with a pitch. We let their specialists try it on a real question and check the answer against the sources. In an institution like that, the person who decides is rarely the person who does the research, so the product has to win the researcher first. Once a specialist has run a cross-border question in an afternoon that used to take a week, and clicked through to every source, the conversation with procurement becomes a formality.<\/p>\n\n\n\n
The formality still has to be done properly. These organisations ask the right hard questions: where is the data hosted, who has access, what happens to our queries, is every source attributable. Moonlit is ISO 27001 certified and GDPR compliant, offers both EU and US hosting and we do not store queries. For a central bank or a Big Four firm, that last point matters as much as the first three.<\/p>\n\n\n\n
What we learned is that the established players are not conservative about technology. They are conservative about risk, which is a different thing.<\/p>\n\n\n\n
We were building the data. For a fund that invests in AI infrastructure rather than AI wrappers, that was the conversation they wanted to have.<\/p>\n\n\n\n
The funding pays for engineering, because connecting dozens of official portals and keeping them current every day is a team’s full-time work, not a project. It pays for Commentaries, which needs paid expert editors before a single page is read. It also lets us build the business in the right order: the data layer first, the products on top of it second, and the market reach third. Without outside money we would have had to sell harder earlier and build less.<\/p>\n\n\n\n
In the Netherlands about 40 cents of every euro you earn goes to the state, and part of that funds the legal system. Much of the commentary that explains the law is written by academics at publicly funded universities. The large publishers sell that commentary back to the universities, the courts, the ministries and the firms that need it, at hundreds of euros per user per month. Everyone who has to comply with the law pays twice. A fair legal system should not cost more per month than your Netflix subscription.<\/p>\n\n\n\n
Commentaries is our answer. Each commentary is drafted from the primary sources in our Data Layer by our own engine.<\/p>\n\n\n\n
The business works because the commentary is not the core product. The Data Layer is. Commentaries is included in the Platform subscription for professionals who research daily, and it reaches AI assistants through the MCP connector and other legal AI products through the API. Every reader who comes for a free commentary sees what sits underneath it, and the ones who need the full platform, or the data behind it, pay for that. The open commentary also gets better with use: readers correct it, editors revise it, and the Data Layer flags every new ruling. A closed commentary cannot do that. So the thing we give away is also the thing that makes the paid product stronger.<\/p>\n\n\n\n
The biggest obstacle was not the industry. It was the publishers’ head start. For decades the knowledge that explains the law has belonged to two or three companies, and the whole market is shaped around that: the subscriptions, the citation habits. When you tell a lawyer the interpretation of the law can be open, free and checked at the source, the first reaction is not disagreement. It is disbelief that it can be done at the quality they are used to.<\/p>\n\n\n\n
You cannot win that debate in a meeting, so we built the proof instead. First the Data Layer, with every document linked to its official source, so nobody has to take our word for the law. Then Commentaries, with named editors from universities, courts, government and law firms.<\/p>\n\n\n\n
The ambition is to be the legal data infrastructure and publisher built for humans and AI. Publisher in the full sense: we collect the law from the official sources, structure and connect it, and now publish the knowledge on top of it.<\/p>\n\n\n\n
The models will keep getting better and the difference will come from what they are allowed to read and the proprietary data you build on top of them. Our own assistant runs on the same MCP connector that outside agents use, and customers already run their own agents on it. As more of the work shifts to reading, the value of a verified layer underneath goes up.<\/p>\n\n\n\n
The second is law firms building in-house. Most of our API customers today are legal tech vendors. The next wave is firms and legal departments that build their own tools on the same data, and the MCP connector lowers the bar to a few clicks.<\/p>\n\n\n\n
The third is geography. The problem we solve is cross-border by definition, so the market is wherever the law is published and people have to comply with it. We host in the EU and the US, and the Data Layer grows one official portal at a time.<\/p>\n\n\n\n
First, build the foundation before the feature. Everyone could see the chat interface coming. Few wanted to spend years connecting official portals and resolving citations, because that work is invisible. It is also what cannot be copied in a quarter. Pick the hard, unglamorous layer in your market and own it. The products on top come easier than you think.<\/p>\n\n\n\n
Second, let the users decide before you do. Moonlit became a company because specialists kept using the internal tool without being told to and kept asking for more. When people change how they work for your product, you have something.<\/p>\n\n\n\n
Third, make every claim checkable. In our market that means a link to the official source under every answer. In yours it will mean something else, but the principle holds: the fastest way through a careful industry is to remove the need for trust. A customer who can verify your work for themselves does not need a reference call, and a sceptic who can check you has nothing left to argue with. Execution over plans, and proof over pitch.<\/p>\n\n\n\n
Image Credits: private picture<\/p>\n\n\n\n