Repertorium · Legal knowledge infrastructure for AI
Law connected. Not just searched.
For people, for applications and for AI agents. Case law, statutes, citations, amendment history and the links between Polish and European law.
Repertorium is a corpus of Polish and EU law prepared for AI models: 1,523,697 Polish law documents, 247,954 European Union law documents and 1,535,990 links between them. Every passage comes back with a locator, that is a document identifier and a character offset, and a citation that could not be resolved stays in the answer with a named reason.
1,523,697Polish documents
247,954EU documents
1,535,990PL-EU links
Try it now · no account · no email · no card · public console since
Explore Repertorium
Try:
Ask it the way you would ask a colleague down the corridor. This searches the full corpus, not a sample.
Test release · EU and Polish corpus · up to a hundred queries a day
We keep query text for 90 days to improve the search. Do not enter client data or case details. Privacy policy
You are searching the full corpus, with no account and no card. Need it inside your product or your agent - write to us and we will scope it.
Repertorium knows the amendment history and can flag that a provision cited in a judgment was amended later.
That is a warning for the lawyer, not a conclusion drawn in their place.
judgment of 2011-03-17
cites: art. 415 Civil Codeamendments_since: 3
status: check the basis
Lifespan
A judgment does not end on the day it is published.
Repertorium lets you watch whether an argument is cited later, developed further, or fades out.
A citation count alone does not show that: a thousand references from twenty years ago and twenty from last year are two different stories.
Repertorium is not a legal chatbot.
It does not write opinions and it does not replace a lawyer. It gives people and AI the sources, relations and context needed to work with law.
A model can generate an answer. Repertorium is there to show what that answer stands on.
Four ways in
One infrastructure. Four ways to use it.
Use it.
Search here, or plug Repertorium into the tool you already work in.
WWW, MCP, your own application.
Train with it.
Use the structured corpus of law for RAG, evaluation, fine-tuning or training systems of your own.
Brand it.
Build your own LegalTech product on the Repertorium data layer.
White label, your interface, your product.
Own it.
Run the part of the corpus and infrastructure you need inside your own environment.
Access for machines
Law available to agents.
Repertorium has a remote MCP interface. It already works with Claude and can serve as a source of law for your own applications and AI agents.
No downloading the whole database and no local copy of millions of documents to maintain.
Confidentiality instruction for Claude
Paste it once. It asks Claude to keep client data out of queries sent to Repertorium and other legal databases.
When you use Repertorium or any other legal-database connector, ask only about legislation, case law and legal questions. Do not put first names, surnames, company names, personal ID numbers, tax or company register numbers, addresses, dates or amounts from the matter into the queries, and do not include case numbers of matters I handle (case numbers of published judgments are fine). Describe facts in general terms (a limited liability company, an employee, a commercial lease). When verifying citations, send only the list of citations, not the text of the document.
Where to paste it
Claude Code - a CLAUDE.md file in your project folder, or ~/.claude/CLAUDE.md for all projects.
Claude app - Settings, Profile tab, personal preferences field.
Claude project - project instructions.
The instruction is a request to the model, not a technical block. What we store and for how long: Where your data is.
Use a corpus edition, white label, or an installation in your own infrastructure.
An account appears only when it is actually needed. The terms each of these routes runs on: Repertorium License.
In the ecosystem
Repertorium stands on its own. It can power your system.
Today PATRON reaches Polish and EU legal sources through its own connectors; connecting it to Repertorium is planned. Repertorium's infrastructure can also run inside other applications and environments.
The console on this page searches the full corpus and asks for nothing. An account comes in only with MCP access, corpus editions and local installation.
Where does the data come from?
From official gazettes and public case law databases for each jurisdiction. Every document carries a source identifier, and the corpus is a snapshot of the law on a given day, not a live stream.
Can I train a model on this corpus?
Training rights sit in a separate licence, because the scope of permitted use differs between jurisdictions. The terms are in the Repertorium License. Tell us what you are building and we will come back with scope.
How is this different from an ordinary case law search?
A search engine finds the document. Repertorium shows its place in the law: links to statutes and other judgments, the amendment history of the legal basis, and whether a line of case law is still alive.
Your AI does not need another model. It needs better access to the law.