← Back to ReadArc

READARC / PRIVACY

ReadArc Privacy Policy

Updated: 2026-11-13

This policy covers the ReadArc desktop app (including the Mac App Store edition and other desktop distributions). ReadArc is a local-first paper reader with no ReadArc-operated accounts, cloud sync, advertising, telemetry or model-request relay. You choose the model provider. When you use cloud AI, the content needed for the task is sent to the endpoint you configure.

Data stored on your device

Imported PDF copies, extracted text, images and thumbnails, reading progress, highlights, translations, summaries, conversations, model usage and settings are stored locally. Reading notes are Markdown files. API keys are stored in a local configuration directory's .env file with restricted file permissions; this is not an encrypted vault. The main configuration stores the keys' environment-variable names.

“Settings → Data” shows the actual storage locations. The Mac App Store edition uses its app sandbox container; non-sandboxed desktop editions default to ~/Documents/ReadArc/Notes for notes and ~/.readarc for configuration. Backup software, sync services or custom folders you configure may copy these files separately.

Network requests and recipients

Feature Recipient Data sent
AI translation, outline translation, questions, summaries, generated notes and cross-paper analysis The model provider or custom endpoint selected for that task Required paper text, titles, context, questions, conversation history and explanation-style settings; image questions also include the selected screenshot
Translated titles in the library The model provider or custom endpoint selected for translation Paper titles only, used to show translated titles in your library; nothing is sent if no translation model is set up
Double-click word lookup The bundled offline dictionary No network request
Translating selected text The bundled dictionary for matching words; otherwise the selected model The selected text when a model is called
Paper search arXiv, Semantic Scholar and OpenAlex; the selected model is also used when rewriting a Chinese query into English Search terms or rewritten keywords
Importing search results The paper's source website A PDF download request
arXiv HTML versions and paper details arXiv (arxiv.org) The paper's arXiv ID (read from the arXiv stamp on the PDF's first page), to download the official HTML version and its images and to fill in the title, authors and year in your library; no paper content is sent. Downloads are cached on your device for offline reading. The Bilingual and Translation views can switch to “PDF” in the toolbar
Automatic paper details arXiv, Semantic Scholar and OpenAlex After import and at startup, for papers missing authors or year: the arXiv ID or DOI printed on the first page; if there is neither, the paper title is sent to OpenAlex (can be turned off in Settings → General). Paper content is not sent
“Fill from ID” when editing paper info Semantic Scholar, OpenAlex or arXiv The DOI or arXiv ID you entered
Listing models and detecting local models Configured provider endpoints, or local Ollama / LM Studio addresses Model-list requests, with an API key when the provider requires authentication
Testing a proxy The configured proxy and Google's www.gstatic.com/generate_204 A connectivity request without paper content or a model API key
Opening project, support or provider links Websites opened in your system browser Normal browser requests

Cloud model requests authenticate with your configured API key. Network recipients can generally also see the connection's source IP address and request time. Requests use a proxy if you configure one. You can inspect and change endpoints in “Settings → Connect”. A localhost address alone does not guarantee local processing: a local gateway may forward requests to cloud services.

PDF parsing and the bundled layout model run locally and do not require uploading an entire PDF to a ReadArc server.

Third parties and your choices

ReadArc does not sell user data or use advertising identifiers for tracking. We do not operate your selected model provider, paper websites or proxies, and cannot determine their retention periods, training use or processing locations. Review their privacy policies and account settings before use. Do not send personal, confidential or restricted material to services that are not authorized to receive it.

You can use offline reading, the bundled dictionary, highlights and notes, or connect a model that actually runs locally. Stop using cloud AI, change the endpoint or remove its configuration to prevent subsequent tasks being sent to that service. These actions cannot recall requests already sent.

Retention and deletion

Local data generally remains until you clear it. “Settings → Data” lets you clear search caches, translations, summaries, conversations, images (including the cached arXiv HTML versions) and usage records separately. Deleting a paper does not delete its Markdown notes.

“Reset all” removes the app database, imported PDF copies, image caches and interface settings, but preserves model configuration, API keys and notes. To remove everything, note these folders in Settings, quit the app and manually delete the retained files. Revoke keys with the provider when necessary. Uninstalling the app or clearing local data does not delete content received by providers or backups you created. Contact the relevant service for remote deletion requests.

Contact and changes

For privacy or security questions, see ReadArc Support. Do not include personal material, API keys or databases in screenshots, logs or attachments.

We update this policy and its date when data handling changes. The app includes an offline copy for its version; the policy published on our website may be updated with newer releases.