Privacy & Technical Guarantees
Your data never leaves your device.
Last updated: August 2026
Parsing happens in RAM
Files are read with the browser File API and parsed with JSON.parse and JSZip. ZIP archives are decompressed inside your browser heap — a 500 MB export never touches a server, a disk, or a network cable.
Background threads, your CPU
Statistics, search and scanning run on a Web Worker. Worker memory belongs to your tab, not to any service. The main thread stays responsive even on multi-gigabyte archives.
Zero server telemetry
The analyzer performs no fetch, no XHR, no beacon, no WebSocket for analytics. There is no tracking cookie, no session ID, no fingerprinting of the analyzer itself. Nothing about your files is ever transmitted.
No account required
The analyzer works without sign-in. Optional Google Sign-In is used only to label support tiers — you can analyze your exports completely anonymously.
Browser sandboxing
The page runs under the browser’s same-origin and script sandbox. JavaScript has no access to other tabs, your clipboard beyond paste, or your local filesystem beyond the files you explicitly select.
Volatile by design
Conversation data lives only in heap memory. We never write your conversation contents to storage, servers, or logs. Close the tab and it is gone permanently.
Advertising, cookies & third-party trackers
The analyzer itself sets no cookies and no server-side telemetry. To keep the project free, we may display advertising served by Google AdSense on public marketing pages, including these help and guide pages. This section describes exactly how that works under the IAB Transparency & Consent Framework (TCF) v2.3.
When AdSense is active on a page, Google may use cookies or similar identifiers to serve ads based on pages you have visited, measure ad performance, and protect against fraud. Google may process your IP address, an anonymous ad identifier, a unique device identifier, and the content a page was viewed by. AdSense cookies are only set on pages where an ad unit is in place — never on the analyzer upload experience itself.
- Personalized advertising is only shown where you have given consent. We rely on a TCF v2.3-compliant consent management platform to collect and store that choice before any ad requests are made.
- Non-personalized ads may still be served without consent. These use contextual signals only and do not create a browsing profile.
- Your choices are yours. You can withdraw or change consent at any time via the privacy controls, reset or delete ad cookies in your browser, and use Google Ads Settings to control personalized advertising across Google services.
Google's AdSense terms require this disclosure. For more detail on how Google uses data when you consent, read Google's "How Google uses information from sites or apps that use our services" .
Concerned your chat data could leak through ad code? It cannot — ads render in an isolated third-party frame with no access to the analyzer's heap, Workers, or Blob URLs. Marketing pages never receive file data.
What we collect, and what we never see
| Data | Handled | Who sees it |
|---|---|---|
| Conversation contents | Client RAM only | Only you |
| Uploaded files (ZIP/JSON/JSONL) | Client RAM only | Only you |
| Search queries & scan patterns | Client RAM only | Only you |
| Exported files (.md / .txt / .json) | Generated locally | Only you |
| Google profile (name, email, avatar) | Optional, on sign-in | You + Google (per their policy) |
| Payment details (support tier) | Optional, via Razorpay | You + Razorpay (per their policy) |
| Ad cookies / identifiers | Marketing pages only | Google (per AdSense terms) |
The browser security model, briefly
Browsers are deliberately hostile to untrusted pages. Scripts run inside a per-origin sandbox: they cannot read
your filesystem, keylog other apps, or access other tabs. The only way this page touches your data is the File object you hand it, and even then it receives a handle, not a path.
That is a hard guarantee of the web platform itself.
Because we never register service workers, never set functional cookies, and never write conversation contents to persistent storage, there is no residue to clean up. Your history exists for the duration of the parse, is held in a worker's heap, and is garbage-collected when the tab closes.
Hosting is static on Cloudflare Pages with strict security headers — X-Frame-Options: DENY, X-Content-Type-Options: nosniff, and a locked-down permissions policy.
The analyzer ships no analytics at all.
FAQ — privacy
Can you see my files?
No. Files are read by your browser, parsed by your CPU, and held in your RAM. The hosting network carries a static HTML/JS bundle and nothing else.
Does it work offline?
Yes. Once the page has loaded, parsing, scanning, search and export all run locally. Disconnect your network and it still works.
Does AdSense see my chat data?
No. Ads render inside an isolated iframe with no access to the analyzer’s memory, Workers, or generated Blob downloads. Ad servers only see page-level signals, never conversation contents.
Is this affiliated with OpenAI or Anthropic?
No. This is an independent tool from UtilityBox. ChatGPT and Claude are trademarks of their respective owners.
What about attachments inside my export?
Anything bundled in the archive — images, PDFs, artifacts — stays on your device. We only decompress and index the JSON conversation data.
Questions about this policy? Email privacy@utilitybox.app . For GDPR or CCPA data-subject requests, the same address applies; because we hold no server-side data about you, such requests are typically satisfied by simply clearing your browser storage.