About Tesseum
A legal engine that cannot invent the law.
Every answer Tesseum produces is bound to an article it has actually read. If the article is not in the corpus, Tesseum will say so — and stop.
Where the idea comes from
A book, then an engine.
Tesseum did not begin as software. It began as a thesis — Epigénesis Biológica del Derecho, "the biological epigenesis of law" — arguing that a legal system behaves less like a frozen code and more like a living one: a hierarchy that inherits, expresses, and adapts, the way a genome does.
That argument is not a metaphor bolted onto a product. It is the architecture. The corpus is a strict hierarchy because the law is a hierarchy. The engine retrieves provisions instead of generating them because, in a living system, what matters is what is actually expressed — not what could plausibly be invented. The book is the specification; the engine is its implementation.
The problem
Legal AI hallucinates. So we built one that cannot.
General-purpose chatbots confidently cite statutes that do not exist. They invent court decisions. They paraphrase precedent that contradicts itself. For research, that is annoying. For a brief filed in court, it is malpractice.
Tesseum starts from the opposite end. The legal corpus comes first — every article, statute, and tesis pre-indexed and verifiable. The AI does not generate citations; it retrieves them. Every claim points back to a passage you can click, read, and verify.
How we are built
Four principles, hard-coded into the system.
Verified retrieval, not free generation
Magnus and Praetor cannot cite an article they did not read. The Praetor quality gate inspects every output and flags unverified references before they reach the user.
Cross-jurisdictional from day one
The corpus is indexed against English-language summaries, allowing a Czech query to find a Mexican statute and vice versa. Same workspace, four UI languages, one verified source.
Source-language fidelity
Legal text stays in its original language. We do not translate statutes — we translate the interface. A Mexican attorney reads Mexican law in Spanish; a Czech attorney reads Czech law in Czech.
European-grade data handling
GDPR-compliant by design. Documents are anonymized before any AI engine sees them. Subprocessor list, DPA, and SCCs are public — no surprises, no audit-day scramble.
Where Tesseum works
Wherever the law is written.
Continental Europe. Latin America. Any jurisdiction whose statutes we verify against the official gazette. Describe your case in any language — Magnus reads every corpus we cover and grows as we add more.
How it works
Your question meets the law.
On the left, the legal corpus — ordered hierarchy of codes, articles and paragraphs. On the right, your question — unstructured, in natural language. Tesseum connects the two in real time, returning the exact provision that applies.
Who we serve
Built for the attorney who cannot afford to be wrong.
Tesseum is designed for independent lawyers, boutique firms, and in-house counsel who need answers fast but cannot tolerate a fabricated citation in a court filing. Law students and academic researchers use the Free tier to learn the corpus; paying professionals use Magnus and Praetor as the working surface for casework.
We are deliberately not building a "general legal chatbot." We are building the working tool for someone who will sign their name to the brief.
The team
Small, and accountable for every citation.
Tesseum is built by a small, independent team working from Prague — and deliberately not a large organisation. Formal incorporation as a Czech limited company (s.r.o.) is in progress; until it completes, the people who build Tesseum stand behind every line of it directly. The subprocessor list, the DPA and the security posture are public, so nothing about how it runs has to be taken on faith.
We are not trying to be everything to everyone. One verified corpus, two engines that read it, and the discipline to say "not in the corpus" instead of guessing. That is the whole of it — and for the lawyer who signs the brief, it is enough.
See it for yourself.
Free tier. No credit card. Verifiable citations from the first query.
Try Magnus View pricing