How File Maktab handles PDFs, images, text files, optional accounts, and translation tools.
Many PDF, image, and text-file tools process the selected file in your browser and create the result on your device. This applies only to tools and operations implemented as local; it is not a promise that every file tool is local. The app may separately send limited operational telemetry to File Maktab (such as route/tool name, event type, duration, or an error code); the client telemetry payload does not include file contents.
In Translate PDF, readable text is extracted from the PDF in your browser and shown for review. When you submit it for translation, the text (not the original PDF or its page images) is sent to File Maktab's server, which forwards it to the configured language-model endpoint and returns translated text. Sign-in is required to submit a translation. A scanned PDF without selectable text may need OCR first.
You can optionally save a completed tool activity to a signed-in account; an activity record contains a tool ID/path, success status, timestamp, and coarse file-size bucket, not a filename, file content, document text, or result download. Favorites and recent tools stay in this browser unless you choose to sync them; synced preferences contain tool titles and internal paths. You can clear activity or delete the File Maktab account in account settings. AI usage counters are also maintained; the repository does not specify a general retention period for account records.
The searchable-PDF OCR tools (OCR PDF and PDF Editor) use Tesseract.js in your browser: the PDF is read, its pages are rendered, and OCR results are assembled locally. Separately, the AI Tools image-to-text feature sends the selected JPG, PNG, or WebP image to File Maktab's server when you run it; the server forwards the image to the configured language-model endpoint and returns extracted text. The repository does not establish how long that submitted image or resulting text may be retained by production infrastructure or the provider.
After Google AdSense approval, free pages may show ads served by Google. Google may use cookies or similar technologies to serve and measure ads; the ads policy page explains this in more detail. Advertisers never receive your files, filenames, or document contents.
You can clear your optional account activity or permanently delete your local File Maktab account from account settings. Local deletion removes the File Maktab user record and account metadata, but cannot delete an external identity-provider account. Delete results and downloads from your device after use, and never upload passwords or API keys to a service you do not trust.
When you submit text for AI features such as summaries, document chat, translation, analysis, or information extraction, that text (and, for chat, the submitted question/history) is sent to File Maktab's server and forwarded to its configured language-model endpoint. Image-to-text sends the selected image as described above. The repository's default endpoint is forge.manus.im, but a production configuration can override it; the deployed endpoint and underlying provider cannot be confirmed from this source. These are server-assisted requests, not local processing.
The reviewed AI procedures return provider responses and update account-level request counters; the source does not show submitted prompts, images, or responses being written to the account activity or usage-counter records. That does not verify logging or retention elsewhere in the production service, at the configured endpoint, or by its underlying provider. The repository does not establish whether production requests or responses are retained, for how long, or whether they are used for training or later deleted. We cannot promise that server-submitted content is discarded after a response; check the applicable production/provider terms and avoid submitting sensitive content unless you have confirmed the protections you need.