[Leer en español](README.es.md)

IureOCR icon

IureOCR

Turn scans and photos of documents into searchable PDFs, on your own computer.

Latest release Downloads License Platforms: Linux, Windows, macOS

Download for Linux · Windows · macOS

A three-page scanned lease goes through OCR page by page in the queue, then its recognized text opens and a clause is selected

IureOCR recognizes the text in scanned PDFs and photos of documents with [Tesseract](https://github.com/tesseract-ocr/tesseract) and produces a PDF with searchable text, ready to upload to [Iurefficient](https://iurefficient.com). It is one of the [Iurefficient desktop apps](#part-of-the-iurefficient-suite) and shares the account, keychain and session with them. ## Why IureOCR - **Your documents stay on your computer.** OCR runs locally with Tesseract; neither the document nor the text is sent anywhere. An Iurefficient account is only needed to save the result there. - **The PDF still looks like the original.** The result keeps the scanned image and adds an invisible text layer on top, so you can search and copy, plus a `.txt` with the plain text. - **Whole case files in one go.** A queue with per-page progress, and PDFs that already have text are skipped instead of being processed again. - **Page tools included.** View, remove, keep, rotate and reorder pages, redact what must not be shared and add Bates numbers, without opening another program. - **Works with your AI assistant.** `IureOCR --mcp` lets Claude, GitHub Copilot and other MCP clients run OCR and read PDFs on this computer, with no OpenRouter key. ## Screenshots
Queue with files at different stages
The queue: one file being recognized (page 6 of 14), one queued, one done and saved to Iurefficient, one that already had text.		 Recognized text of a scanned lease
The recognized text of a finished document, ready to select and copy.
Page grid with three pages selected
Page tools: mark pages to rotate, keep or remove them, or drag them into a new order.		 Settings, AI assistants section, in dark theme
Settings → AI assistants: connect Microsoft 365 Copilot, GitHub Copilot in VS Code or Claude Desktop.
## Features - **Local OCR with Tesseract** in Spanish and English (more languages if your Tesseract has them). Nothing leaves the computer: neither the document nor the text. - Accepts **scanned PDFs** and **images** (JPG, PNG, TIFF, BMP, WEBP). The result is a PDF with the original image and an invisible text layer on top: it can be searched and copied, and the Iurefficient server indexes it without processing it again. It also leaves a `.txt` with the plain text. - **Skips PDFs that already have text** (same rule as the server: fewer than 100 characters in the first three pages = needs recognition). It can be forced. - **Queue** with per-page progress, cancel and retry. Opening a file does not start OCR by itself: it stays as "No OCR" so you can view it, edit its pages or upload it, and **Recognize text** (or **Recognize text in N** for all of them) starts it. Settings has an option to recognize files as soon as they are opened. - **Built-in viewer**: continuous scroll, go to page, zoom (buttons, `+`/`-`, Ctrl+wheel) and keyboard (PgUp, PgDn, Home, End, Esc). Only the visible pages are drawn, so files with hundreds of pages open without waiting. It highlights the page being recognized and shows the resulting PDF when OCR finishes. - **Page editing** with thumbnails: remove pages, keep only the selected ones, rotate them (90° right by default; 180° and 90° left from the menu), or reorder them by dragging (also "To start", "To end" and "Reverse order"). Edits pile up in a temporary copy until you choose **Save as copy** (` - edited.pdf` next to the original) or **Replace original**; closing asks before discarding them. - **Redaction.** Mark areas on a page and everything under them is removed for good: the text, including the invisible OCR layer, and the pixels of the scanned image. The result is verified, and nothing is written if anything is left under a mark. - **Bates numbering** on every page, with prefix, suffix, digits, start number and one of six header or footer positions. - Add files by dragging them onto the window, opening them with IureOCR from the system or with `iureocr://ocr?file=…` links. - **Save to Iurefficient**: to a project (REST API) or to any folder in the document tree (WebDAV). With the IureDav drive mounted, just choose it as the output folder. - For editing the result it suggests **OnlyOffice Desktop Editors**: it detects it if installed and otherwise links to its download. It is not bundled. - **English and Spanish interface**: English by default, and Spanish if the operating system is set to Spanish; it can be changed in Settings → Appearance. This is independent of the text languages Tesseract recognizes. Planned for upcoming versions (see the [CHANGELOG](CHANGELOG.md)): merging PDFs, splitting by ranges, password protection, watermark and page numbers; real image compression; rescuing unreadable pages with a vision model; the instance's list of documents pending OCR; and Tesseract bundled on macOS too. ## Download and install Get the installer from the [latest release](https://github.com/ellaguno/iureocr/releases/latest): | File | System | | --- | --- | | `IureOCR__1-windows-x64.exe` (or `.msi`) | Windows 10/11, x64. Tesseract is included. | | `IureOCR__2-macos-universal.dmg` | macOS 12 or later, Intel and Apple Silicon. Needs `brew install tesseract`. | | `IureOCR__3-linux-x64.deb` / `.rpm` | Debian/Ubuntu or Fedora. They install Tesseract with Spanish and English as dependencies. | | `IureOCR__3-linux-x64.AppImage` | Any other Linux distribution. Install Tesseract from your distribution first. | The installers are not code-signed, so Windows and macOS warn the first time you open them; see [Code signing policy](#code-signing-policy) for what that means and how to continue. ### Tesseract on each system | System | Where Tesseract comes from | Language models | | --- | --- | --- | | Linux | the system package: the `.deb`/`.rpm` depend on `tesseract-ocr` with `spa` and `eng` | the ones bundled with the app (`tessdata_fast`) | | Windows | **bundled in the installer** (UB Mannheim build, Apache-2.0) | bundled | | macOS | `brew install tesseract` (not bundled for now) | bundled | The app looks for Tesseract in this order: the `IUREOCR_TESSERACT` variable, the one inside the installer, the usual system paths and the `PATH`. Settings shows which one it found, its version and the available languages. ## How it works inside 1. **pdfium** rasterizes each PDF page to JPEG at the chosen resolution (300 dpi by default). Images are passed as they are. 2. **Tesseract** receives the list of pages and produces the PDF with the text layer itself (its `pdf` renderer, the same one ocrmypdf uses) and the `.txt`. 3. The result is saved next to the original with the suffix ` - OCR` (configurable) or in a fixed folder. A PDF that already has readable text is not rasterized: the app says so and offers to force it. ## Use from Claude, Copilot and other agents (MCP) `IureOCR --mcp` starts an [MCP](https://modelcontextprotocol.io) server over stdio, with no window. Claude Desktop, Claude Code, Copilot in VS Code or any MCP client can then use the OCR and PDF tools on this computer. OCR runs here and the client's model reads the text and does the rest, so no OpenRouter key is needed. Only the text the agent chooses to read leaves the computer. | Tool | What it does | | --- | --- | | `ocr_file` | OCR a scanned PDF or an image; writes the searchable PDF and the `.txt` next to the original and returns the text | | `extract_text` | Reads a PDF's text layer by page range (or a `.txt`), for long documents | | `pdf_info` | Pages, sizes and whether the PDF already has text | | `render_page` | A page as an image, so the model can see signatures, stamps or tables | | `edit_pdf_pages` | Remove, keep, rotate or reorder pages (new file) | | `ocr_languages` | Tesseract version and installed languages | The original is never modified and paths must be absolute. Languages, resolution and output folder come from the app's Settings. The log goes to `iureocr-mcp.log`, next to the app's. `IureOCR --mcp-config` prints what to paste into the client's configuration, with the executable's real path: ```json { "mcpServers": { "iureocr": { "command": "/path/to/IureOCR", "args": ["--mcp"] } } } ``` In IureOCR, **Settings → AI assistants** detects which assistants are on the computer and connects each one the way that works for it: - **Microsoft 365 Copilot** (the one almost everybody has) cannot use programs on the computer, but it can use the Iurefficient instance's MCP server. While signed in, **Create access for Copilot** generates an `iurmcp_…` token (valid for a year) and shows the server URL and the steps to add it to an agent in Copilot Studio (`X-MCP-Token` header). Copilot sees what the user sees in Iurefficient; to let it read a scan, save the OCR result there. Tokens are revoked from the same row. - **GitHub Copilot in VS Code**: **Connect** opens the `vscode:mcp/install?…` link; VS Code asks for confirmation and saves the server in its settings. IureOCR does not write them. - **Claude Desktop**: **Connect** adds the entry to `claude_desktop_config.json` without touching anything else (a `.bak-iureocr` copy is kept) and warns if the app has moved. Then quit and reopen Claude Desktop. Other clients: `claude mcp add iureocr -- /path/to/IureOCR --mcp` in Claude Code, or the `--mcp-config` entry in their configuration. ## Part of the Iurefficient suite | App | What it does | | --- | --- | | [IureTranscribe](https://github.com/ellaguno/iuretranscribe) | Local Whisper transcription, live recording with who-spoke, summaries and minutes. | | [iureditor](https://github.com/ellaguno/iureditor) | WYSIWYG Markdown editor with Mermaid, LaTeX and PDF/DOCX export. | | [IureDav](https://github.com/ellaguno/iuredav) | Mount a WebDAV server (or Iurefficient) as a drive. | | **IureOCR** | Local OCR that turns scans into searchable PDFs. | | [iureTI](https://github.com/ellaguno/iureTI) | IT asset discovery probe for the Iurefficient inventory. | ## Contributing Issues and pull requests are welcome. Good first contributions: - **Bug reports with the log file** attached (see [Where it stores things](#where-it-stores-things) for its location), and the system and version you use. - **Translations**: a new interface language is a new dictionary in `src/lib/i18n.svelte.ts`, next to the English and Spanish ones. - **Documentation**: corrections and clearer explanations in this README. To build and run the app, see [Development](#development) below. ## Development
Requirements, building and checks Requirements: Node 22, stable Rust, the Tauri 2 dependencies for your system and Tesseract installed (on Linux `sudo apt install tesseract-ocr tesseract-ocr-spa tesseract-ocr-eng`). ```bash npm ci python3 scripts/descargar-pdfium.py # pdfium library for this platform python3 scripts/descargar-tessdata.py # spa, eng and osd models (tessdata_fast) npm run tauri dev ``` On Windows, `python3 scripts/descargar-tesseract-windows.py` also fetches the Tesseract that gets bundled (requires `7z`). ```bash npm run check # frontend types cd src-tauri && cargo check # backend npm run tauri build # installers for this platform ``` The screenshots and GIFs in this README are generated from the real interface with a mocked backend: see [`scripts/readme-media/`](scripts/readme-media/README.md).
Releasing a version Bump the version in `package.json`, `src-tauri/Cargo.toml` and `src-tauri/tauri.conf.json`, add the section to the CHANGELOG and tag: ```bash git commit -am "v0.2.0: …" git tag -a v0.2.0 -m "IureOCR 0.2.0" && git push --follow-tags ``` `.github/workflows/build.yml` builds on every push and, with a `vX.Y.Z` tag, publishes the release with installers for Windows (`1-windows-x64`), macOS (`2-macos-universal`) and Linux (`3-linux-x64`, `.deb`, `.rpm` and AppImage), plus the updater's `latest.json`. Updates are signed with the app's minisign key (`~/.tauri/iureocr-updater.key`). The workflow also has Authenticode signing steps through SignPath, which only run when the `SIGNPATH_*` secret and variables exist; they are not configured, so Windows installers are published unsigned.
Updates At start-up (it can be turned off in Settings) and with **Check now**, the app looks for a newer release. The Tauri updater installs it from inside the app and verifies its minisign signature first. Each installation receives its own kind of package: the `.exe` on Windows, the `.app.tar.gz` on macOS, the AppImage, or the `.deb`/`.rpm` (installing one asks for the administrator password). If the update cannot be installed, the app shows the error and a button to open the download page.
### Where it stores things | What | Linux | Windows | macOS | | --- | --- | --- | --- | | Settings | `~/.config/com.iurefficient.iureocr/settings.json` | `%APPDATA%\com.iurefficient.iureocr\settings.json` | `~/Library/Application Support/com.iurefficient.iureocr/settings.json` | | Log | `~/.local/share/com.iurefficient.iureocr/logs/iureocr.log` | `%APPDATA%\com.iurefficient.iureocr\logs\iureocr.log` | `~/Library/Application Support/com.iurefficient.iureocr/logs/iureocr.log` | Credentials go to the system keychain, shared with the other Iurefficient apps; the active account (instance and email) lives in `/iurefficient/`. ## Code signing policy - **Windows installers are not code-signed for now.** The first time you run one, SmartScreen may show "Windows protected your PC" with an unknown publisher: choose **More info → Run anyway**. - **macOS: the app is not notarized** by Apple. The first time, right-click (or Control-click) IureOCR in Applications and choose **Open**, or go to **System Settings → Privacy & Security** and click **Open Anyway**. - Every installer is built from this repository by GitHub Actions (`.github/workflows/build.yml`) and published only on the [releases page](https://github.com/ellaguno/iureocr/releases). Download it from there. - What is signed: the update packages (`.exe`, `.msi`, `.app.tar.gz`, AppImage, `.deb` and `.rpm`) carry a minisign signature (`.sig` files and `latest.json`), and the app checks it against the public key it ships with before installing an update. - **Maintainer** (commits and reviews): Eduardo Llaguno ([@ellaguno](https://github.com/ellaguno)). ### Privacy policy This program will not transfer any information to other networked systems unless specifically requested by the user or the person installing or operating it. Text recognition runs entirely on the user's computer with [Tesseract](https://github.com/tesseract-ocr/tesseract); documents never leave the machine. Specifically, IureOCR connects only to: - the Iurefficient instance that the user configures, and only once an account is set up: to check the session (at start-up and in Settings), and when the user signs in, saves a document there or manages the access tokens for Copilot. Without an account it is never contacted; - GitHub (`api.github.com` and the release files on `github.com`): at start-up, to check whether a newer release exists (it can be turned off in Settings), and when Settings is opened, to show the latest version of the other Iurefficient apps. It collects no telemetry and no usage statistics. Credentials are stored in the operating system keychain, never in configuration files. ## License Apache License 2.0 (see [LICENSE](LICENSE)). Tesseract (Apache-2.0), pdfium (BSD-3) and the `tessdata_fast` models (Apache-2.0) keep their own licenses. Redaction and Bates numbering use the engine of [PdfCraft](https://github.com/storytold/pdfcraft) (MIT OR Apache-2.0); see [NOTICE](NOTICE).