Anonimatum 1.8.0 — MCP, password change, more accurate OCR and lighter PDFs

Version 1.8.0 exposes Anonimatum to AI assistants via MCP (= Swagger), lets you change your password from your profile and strengthens security. OCR is only applied, and only charged, on pages that are really scanned, and copyable text is more faithful. It also reviews the generic patterns, produces lighter PDFs and fixes PDF tools and pseudonymization. Release date: 13/10/2026.

MCP connector

• Assistants (ChatGPT, Claude, Cursor, Grok, etc.) can list, describe and run the same API as Swagger.
• Grant and revoke access under Profile → MCP connectors (OAuth 2.1 + PKCE or a manual secret with mcp.json).
• ChatGPT, Claude, Grok and DeepSeek cards with a tutorial: «Connect» opens the assistant's connectors screen and copies the URL; manual access (mcp.json) lasts one year.

Stronger security

• The server no longer trusts identity data sent by the browser: only your session counts.
• Administrator permissions checked exactly and service-to-service routes protected by an internal key.
• Only you can use your patterns, word sets and pseudonymization maps.
• Nothing is anonymized without an active plan, including pseudonymization and AI.
• Technical logs without personal data and timeouts on calls between services.
• Change your password from your profile (Profile → Password): current password, new one and confirmation, following your organization's password policy.

OCR only where needed

• Only suggested on pages that are really scanned: covers with a logo, signatures or blank pages no longer ask for it.
• Only those pages are processed and charged, and the dialog no longer reappears when you change options.
• More faithful searchable PDF: words no longer stick together, € and quotes are kept, the DNI letter is corrected and skewed pages are straightened.
• Recognises Catalan, Galician, Basque, Portuguese, French and Romanian; PaddleOCR 3.7 engine.
• A scanned page is no longer mistaken for a logo in “Images and logos”.
• Detects the Spanish Social Security number (NSS/NAF) and fewer surnames are missed.

OCR engine, copyable text and quota

• Choose Tesseract (fast, default) or PaddleOCR (more accurate, slower) in your profile.
• The searchable PDF is generated in the background; when it finishes, download it from the document queue.
• Quota: document pages + OCR pages. From the API, copyable text also uses quota and accepts specific pages.

Generic patterns and lighter PDFs

• 26 reviewed patterns: DNI/NIE/CIF, full IBAN, phone numbers with prefix and spaces, amounts with thousands separators, NSS, cadastral reference, licence plates, case numbers, IP…, loadable per group.
• Images to PDF, searchable PDF and anonymized PDF use JPEG and take up much less space; the anonymized PDF in high quality keeps images lossless.

New PDF tools and per-plan quota

• Compress PDF (low, medium or high) keeping the text, showing the size before and after.
• PDF to JPG/PNG with page selection, page numbering with preview and PDF metadata (view, edit or remove).
• Thumbnails before processing and a live watermark preview; Word to PDF also accepts OpenDocument and .txt.
• Each plan can include the PDF tools with a monthly usage limit, separate from the anonymization quota and shown on Home and Profile. The warning tells you whether your plan does not include the tools or you have used up this month's quota.

PDF tools

• Electronic signature: AutoFirma opens when you confirm the area; logo or background from the interface.
• Clear error messages and range validation when splitting and reordering.
• Watermark with any character and centred on rotated pages.
• Protect PDF with a separate owner password and AES-256 encryption; wrong-password warning when unlocking.
• PDF to Word converts better and downloads are no longer cut off.
• Page limit per document (500 by default) with a clear warning when it is exceeded.
• PDF to Excel recognises tables: each column in its own cell and amounts as numbers.
• Faster Office conversions (several at once); all tools share the same upload area, messages and button, and prevent double submission.

Pseudonymization: verified output, saved maps and realistic substitutes

• “Image + text” PDF by default: pages as images with an already-replaced text layer (searchable and copyable) and a check that no original data remains; the high-fidelity vector mode is still available.
• Saved maps on the server: the equivalence map is stored encrypted per group and no longer travels in the ZIP; you deanonymize from Saved maps and the administrator decides who may do so.
• The same value gets the same substitute across the whole project, and it cannot be guessed.
• Realistic substitutes per language: names with their surnames, matching emails, valid DNI/NIE, IBAN, phone numbers and licence plates, and addresses with a consistent postcode.
• Method per data type, a list of unique values with editable substitutes and a side-by-side view original ↔ pseudonymized.
• Pseudonymization also removes data hidden inside reused PDF elements (Form XObjects) without turning the page into an image.

Pseudonymization

• In Word, Excel and PowerPoint replacements match whole words (“Ana” no longer changes inside “Mariana”) and are not chained.
• Dates keep their original format and amounts their thousands separator.
• Partial masking is no longer treated as reversible, so a wrong value is never restored.
• Correct gender for full names; substitutes with €, quotes and Romanian characters.
• The resulting PDF removes metadata, annotations, forms, attachments and hidden layers, and is checked so that no original data remains (if any did, it is generated in image mode).
• Rotated margin text (stamps and signatures) is replaced in place and with the same orientation, covering the original.

API keys in your profile and on-premise

• Plans that include it (Enterprise by default) let you create and revoke API keys from your Profile; the full key is only shown when it is created, the list shows them masked and each user can only revoke their own.
• On-premise installation: the same application is configured at startup, with no specific build, and the server sets up the group and plan automatically.

Anonymization

• The × to remove a mark is placed at the end of the text also when the text is rotated or vertical.
• In scanned PDFs, side stamps are confirmed by reading their text: table borders and form columns are no longer marked as stamps.
• Boxes in the right place on invoices with fonts such as Segoe UI and on PDFs exported from Google Docs.
• Fewer AI false positives in technical and audit reports: headings, ISO standards and common words are no longer flagged.
• The AI detects the secure verification code (CSV) and keeps multi-line addresses within their bounds.

Experience and stability

• Warning for files in an unsupported format or larger than 100 MB.
• Reliable downloads from the queue and the result; the queue refreshes without interrupting you.
• No unexpected sign-outs.
• Sidebar collapses on narrow screens and less memory is used when several documents are open.