Back to home

Privacy Policy for MD to DOCX (Offline)

Last updated: September 24, 2026

MD to DOCX (Offline) is a Chrome extension for converting Markdown content into DOCX and PDF documents, importing documents into a Markdown editor, and collecting selected content from supported AI chat websites. Some features run entirely in your browser; others are optional online features that send data to a server only when you use them. This policy explains which is which.

Local conversion, import, and export

Markdown-to-DOCX conversion, PDF export, and ordinary PDF import run locally in the Chrome extension. Importing .md, .txt, .docx, or .pdf files into the Markdown Live Editor is processed in the browser extension runtime and does not require the developer's server.

When importing a PDF, the extension reads the PDF's text layer where possible and uses offline OCR packaged with the extension for scanned or uncertain pages. OCR results may contain errors. Ordinary PDF import and export do not send the complete PDF to the developer backend.

Local PDF import is limited to 100 pages and 25 MiB per file. These are product limits and are not security or privacy guarantees.

Optional Markdown attachment in exported PDFs

When exporting a PDF, you can choose to embed the source Markdown inside the downloaded PDF file. This option is off by default. The embedded Markdown stays inside the PDF you download and can be recovered by the extension if you import that PDF later. This feature does not itself send the Markdown to the developer backend.

Important: anyone you share a PDF with this attachment can extract the source Markdown. Leave the option off if you do not want to share the source text.

Optional PDF AI recovery

PDF AI recovery is an optional online feature offered only when local text extraction or OCR is uncertain. Local PDF import and PDF export continue to work without it.

  • You manually select a problematic region, which must be smaller than 70% of one page.
  • You must expressly consent before each server-assisted recovery request.
  • The extension sends only: a JPEG image of the selected region, the local OCR text for that region, the page number, a SHA-256 hash of the PDF, and task/attempt identifiers.
  • The complete PDF is not uploaded by this workflow.
  • The developer backend forwards the image crop and OCR hint to OpenAI's API for processing.
  • The result is shown to you as a suggestion for review and is not inserted into the editor automatically.

Other optional online AI features

The extension also includes optional online helper features: Humanize, Refine, and AI chain generation. These features are not required to use the converter. When you explicitly use one of them, the extension sends only the content needed for that action to the developer-operated backend at https://maksym342.pythonanywhere.com, which forwards it to OpenAI's API for processing.

Data read on supported AI websites

The extension runs content scripts only on supported AI websites, including ChatGPT, Gemini, Claude, Grok, and Perplexity. On these websites, it reads page content only to provide visible user-facing features such as selection capture, the "Add to MD" action, and the Markdown Toolbox interface.

Data stored locally in Chrome

The extension may store Markdown editor drafts, conversion settings, UI preferences, Toolbox data, a generated extension install identifier, and short-lived ChatGPT token data for user-requested ChatGPT Toolbox features. ChatGPT Toolbox token data is stored locally for that functionality and is not sent to the developer backend.

Local extension data remains on your device until you clear it, reset the extension settings, or uninstall the extension.

Backend storage

The backend stores quota and abuse-prevention metadata in SQLite, including install ID, usage date, daily usage counters, temporary chain ticket IDs, ticket mode, and timestamps.

For PDF AI recovery, the backend additionally stores task and attempt IDs, document and request hashes, token reservations and usage (including reservations where usage is unknown), request statuses, timestamps, and a hashed rate-control scope.

The backend does not intentionally store selected text, generated text, PDF images or crops, OCR text, recovered Markdown, or complete PDFs in SQLite. This content is processed in memory for the request and forwarded to OpenAI. Hosting and service providers may keep their own request logs under their own policies.

Retention

  • Usage counter rows: retained for up to 180 days.
  • Temporary chain ticket rows: retained for up to 7 days.
  • PDF AI task and attempt rows: eligible for cleanup after 30 days.
  • PDF AI rate-control events: eligible for cleanup after about 2 minutes.
  • Monthly token counters: eligible for cleanup after their month ends, unless an attempt is still running.

PDF AI cleanup runs when a later PDF AI request is processed, not at an exact deadline, so eligible rows may remain for some time. Request metadata handled by PythonAnywhere and processing by OpenAI are subject to those providers' applicable policies.

Data sharing

User data is not sold and is not used for advertising.

Data is shared only as needed to provide user-requested features: with the developer backend on PythonAnywhere; with OpenAI through the backend for optional AI processing (Humanize, Refine, AI chain generation, and PDF AI recovery); and with ChatGPT/OpenAI web services when you use ChatGPT-specific Toolbox features in your own browser session.

Human access

The developer does not use submitted text or images for advertising and does not routinely read user-submitted content. Human access to backend data is limited to maintenance, debugging, abuse prevention, security, or legal compliance.

Security

Online requests are sent over HTTPS. The extension does not contain the developer's OpenAI API key.

Contact

For privacy questions, use the support contact provided in the Chrome Web Store listing.