English | 简体中文

MultiTool Office Next

Find files, preview documents, and rename scans in one Windows desktop app.

MultiTool Office Next (MTO) is a free, open-source toolbox for column-based file browsing, local file search, document previews, and offline OCR batch renaming. Browsing, search, preview, and default OCR run on your PC without an app account or AI API key. An AI assistant and translation are available through services you choose to connect.

Download for Windows x64 · Quick start · Features · Report an issue

Current app and installer language: Simplified Chinese.

Installation

1. Open the latest release.
2. Download the Windows installer ending in _x64-setup.exe and run it. You do not need the source-code ZIP.
3. Follow the installation wizard, then open MultiTool Office Next.

Requires Windows x64 and Microsoft Edge WebView2. The installer may download WebView2 if it is missing; OCR models are bundled with the installer. The app window uses the name MultiTool Office.

The app is free to use. Optional model, search, and scoring services may require provider accounts and paid API access. See the release notes for what the downloadable build includes; main may contain newer fixes.

Quick start

1. Add a working folder: in the first-run setup or File Workspace (文件工作区), choose 添加工作目录 and wait for local indexing. You can skip AI setup.
2. Find a file: press Ctrl + F and enter part of its filename.
3. Check its contents: hold Ctrl while hovering over a result to preview it. Double-click to open it in the default application.

You can now browse folder levels, search your selected working folders, and preview files from one window.

Screenshots

Screenshots and GIFs come from existing app builds; some controls have changed in newer versions.

Light mode	Dark mode

Features

File browsing and local search

Keep several folder levels visible while moving through project or client folders.

• Search filenames across your configured working folders; narrow results by file extension, modification date, or path.
• Filter folder names directly in the visible directory levels, within each level’s selected parent. The selected path stays visible; use × to clear the filter and restore the lists.
• Clear file-search keywords with × while retaining other filters.
• Copy names or paths, open files in their default applications, or reveal them in Windows File Explorer.
• Rename one file or folder using the selected-item action or F2. Extensions are preserved by default. Edit the name and press Enter or confirm once; rename history supports persistent records and guarded undo.

Renaming is restricted to ordinary objects inside local working folders and never replaces an existing target. Identity and state checks use Windows file IDs, sizes, and timestamps, without reading document bodies. Case-only changes, links, protected objects, target paths over 240 UTF-16 units (including descendants), and folders exceeding 10,000 descendants are currently rejected. Uncertain interrupted operations require manual review. Running OCR tasks block renaming, and affected OCR results must be imported again.

The index stores filenames, paths, and other metadata. Search does not index document body text.

Local document previews

Hold Ctrl and hover to preview; use Ctrl + left-click or Ctrl + Enter to keep a preview open.

• Preview PDF, common image formats, text, Markdown, CSV, Word (.docx), and Excel (.xlsx).
• Turn PDF pages, switch Excel worksheets, and resize the preview window.
• Preview content stays local and is not automatically added to AI conversations. Macros, formulas, and scripts are not executed.

Word previews focus on text. Excel previews have row, column, and content limits and use cached formula values. Use the original application for complex layouts, charts, and editing.

Offline OCR and batch renaming

Import scanned PDFs or images, extract fields such as company, date, and document number, and assemble filenames using your rules.

Add files → match or select a profile → recognize and rename in one click → view the results.

• Local OCR supports label-based extraction, regular expressions, drawn regions, and date normalization.
• Match profiles per file using first-page keywords. Ambiguous matches or low-quality text stay available for review; change the profile and retry only the affected file.
• Reuse, import, and export profiles, or ask the optional AI service to draft one for review.
• Edit extracted values, inspect their sources, export results, and retry failed or interrupted items.
• The one-click action runs OCR and renames eligible files using the selected profiles, without another rename button or filename confirmation. Optional TXT / CSV / JSON export settings are retained. Uncertain results and errors are held for review; use the OCR-only action for the step-by-step flow.
• Invalid filename characters such as slashes are automatically replaced with underscores without requiring review. Extracted values and exports retain the original text; truncated names and uncertain recognition still require review.
• Renaming checks file state and naming conflicts, reports partial failures, and supports undo. Export an exceptions CSV and retry failed undo operations after resolving file locks or conflicts.

Renaming changes filenames, not document contents. Undo checks whether files have changed or their original names are already occupied. One-click tasks follow the selected profiles' naming rules; uncertain recognition results are held for review.

See the OCR profile and example guide (Chinese) for purchase contracts, quotations, delivery notes, and other forms.

Optional AI assistant

Connect your own model service to ask for file searches, check whether related files are missing, or prepare a document from the results. For example: “List folders that contain Contract.docx but are missing Contract.pdf.”

• Tool calls are shown as the assistant works. Public web search is optional through Tavily or DeepSeek; image questions require a vision-capable model.
• Reading document contents or creating files requires a separately authorized assistant folder and confirmation for each operation.
• Create basic TXT, Markdown, Word, Excel, PowerPoint, HTML, and PDF documents. Outputs focus on text, simple tables, and bullet slides; new files do not overwrite existing ones.
• PDF creation and local preview are supported; reading PDF body text through the assistant is not currently supported.

Extra DeepSeek page fetching is off by default and shares the task's web-request limit with search. AI, translation, and other connected services send relevant information to your chosen providers; see Privacy and data.

Everyday tools

Tasks with a specific date and time can remind you 10 minutes before they are due. The app must remain running; closing the main window can leave it in the system tray.

Optional AI setup

You can reopen the setup guide from 配置引导 in the top bar or from Settings. Local file tools and default OCR work without completing this setup.

Keyboard shortcuts

Privacy and data

File-browser folders and the assistant's authorized read/create folder have separate permissions. API keys are stored in Windows Credential Manager. Conversations, OCR results, tasks, and generated files may remain on your PC until you clear them.

Connected features request consent on first use and when the service, account, or data scope changes. In 隐私与联网 (Privacy and network), revoke consent, enable local-only mode, or inspect activity. Local-only mode blocks application API requests while local file tools and OCR remain available.

Activity keeps metadata and DeepSeek tool-cost snapshots for up to 30 days and 1,000 entries, without additionally recording body text, full file paths, or keys. Setup progress, consent, and activity records are not transferred with exported settings.

Further details (Chinese): Usage and data safety · Assistant folder permissions · Release changes and manual checks.

Troubleshooting

• WebView2 is missing or installation fails: install the runtime using Microsoft's official WebView2 download, then run the app installer again.
• A model or search connection test fails: check the key, account balance, model permissions, endpoint, and network using the reported error.
• Document formatting looks different: previews and AI-created documents have the limits described above. Open the file in its original application for full layout and editing.

Contributing

Bug reports, feature requests, and pull requests are welcome. When opening an issue, include the app version, Windows version, and steps to reproduce. See CHANGELOG.md for release history (Chinese).

git clone https://github.com/d8349565/MultiTool-Office-Next.git
cd MultiTool-Office-Next
npm.cmd ci
npm.cmd run ocr:build
npm.cmd run desktop

License

Apache License 2.0. Copyright 2026 Lee; see NOTICE for attribution. Third-party dependencies, fonts, and icons retain their respective licenses.