Your case files in AI. Without exposing data.
Your case files, ready for AI work. Upload contracts, pleadings and scans: IURA recognizes the text, masks personal data and encrypts every document with your account's key. Ask about them in IURA chat or in Claude, ChatGPT and Cursor through MCP. Data stays in the EU.
From scan to working document
A repository built for case files: scanned rulings, signed contracts, amendments in mixed formats.
Upload
Drag in PDF, DOCX or DOC files, scans included. Organize them in folders by matter or client. Encryption starts the moment a file lands.
Text recognition
Scans become searchable text. The viewer shows the original next to the recognized text, page by page.
Masking
Names, companies, addresses, national IDs, tax numbers and bank accounts are replaced with consistent pseudonyms before the text reaches the model. You see the original.
Ask
In IURA chat or in Claude, ChatGPT, Perplexity and Cursor. Every answer points to the document and page it came from.
The model sees pseudonyms. You see the original.
Masking (pseudonymisation) runs before indexing: the model only ever receives text that has been masked. A preview toggle cannot unmask a document.
Lease agreement
The Landlord John Smith, ID 85010212345, lets to the Tenant Nova Office Ltd the premises.
§ 1. The subject of the lease is the commercial premises at 4 Polna St, Kraków, with an area of 86 m², together with a parking space.
§ 4(2). The deposit equals twice the monthly rent.
Consistent pseudonyms
The same person is always [PERSON_1], across the whole document and every inflected form of the name. The model's answers stay readable.
Manual hiding
Select any passage in the viewer and hide it, or replace it with your own label such as "the Client". A false positive is undone with one click.
"What the model sees" preview
Switch the viewer from the original to the masked version and check exactly what reaches the AI before you ask.
Automatic detection
People, companies, addresses, e-mails, phone numbers, national IDs, tax and company registry numbers, bank accounts, identity documents and case numbers.
Pseudonymisation is not anonymisation. Masked data remains personal data under the GDPR and is protected like the original: the pseudonym mapping is encrypted with your account's key and never reaches the model.
A repository that knows your files
After upload, IURA organizes the documents so that both the agent and you get to the point faster.
- Parties
- J. Nowak / Orion Ltd
- Deadlines
- Notice by 30 Jun 2026
- Amounts
- Rent 4,200 PLN/month, deposit 8,400 PLN
- Open questions
- Does the 2024 amendment change § 12?
Matter folders
Group documents by matter or client. The agent can search a whole folder or a single document.
Type and metadata
Every document gets a type (contract, pleading, ruling, correspondence) and an extract of parties, dates and amounts.
One-pager and folder card
A one-page summary of every document, plus a folder card with notes, priorities and open questions the agent can keep up to date.
Search by meaning
Hybrid, semantic or keyword mode. A result names the document, folder and page; the source opens next to the recognized text.
In IURA chat and in your AI tools
The repository is one of the scopes of the IURA MCP connector. You enable it when connecting and can revoke it at any time.
IURA agent in chat
One switch in the repository settings decides whether the agent may search and cite your documents in conversations.
MCP connector
At authorization you pick the scopes: Polish, Dutch and EU law, and the repository. The agent in Claude searches folders, reads summaries, adds notes and uploads files straight from the conversation.
Always the masked version
In chat and through MCP the model only receives masked text when masking is on. No exceptions for individual tools.
Confidentiality by architecture
A law firm's documents are covered by professional secrecy. The repository is built around that fact.
Per-account encryption
Every document, its text and the pseudonym mapping are encrypted with keys unique to your account. No shared store, no cross-account access.
Data stays in the EU
Storage and processing run on European infrastructure. Documents do not leave the EU.
Never used for training
Your files are your files. They are not used to train models, IURA's or anyone else's.
Access under your control
Documents are visible only to the account that uploaded them. You grant and revoke agent and connector access yourself, at any time.
Frequently asked questions
Are my documents safe?
Yes. Every document is encrypted with keys unique to your account, data is stored and processed in the EU, and your files are never used to train models. With masking on, the model only sees the pseudonymised version.
How do I ask AI about my own documents?
Upload files to the repository. Within minutes they are recognized, masked and indexed. Then ask the IURA agent in chat, or connect the repository through MCP to Claude, ChatGPT, Perplexity or Cursor.
Which file formats does the repository support?
PDF, DOCX and DOC, scans included. OCR recognizes the text, also in scanned and hand-signed documents.
What exactly gets masked?
Automatically: people, companies, addresses, e-mails, phone numbers, national IDs, tax and registry numbers, bank accounts, identity documents and case numbers. Manually: any passage you select in the viewer. Every pseudonym can be reverted.
Is masking the same as anonymisation?
No. It is pseudonymisation: a mapping of pseudonyms exists, but it is encrypted with your key and never reaches the model. Masked data remains personal data under the GDPR and is protected like the original.
How many documents can I upload?
The limit is counted in pages: 1,000 pages on Partner and MCP Basic, 5,000 pages on Senior Partner and MCP Pro, custom for teams. Delete documents to free up space.
Bring your documents into IURA
Try the repository on the platform or talk to the team about document workflows in your firm.