❦ FarziPedia
❦

— Liber de Privatis —

Privacy Policy

What we collect, how we handle it, and what we do not do.

Effective date: 2026-05-25 · Last updated: 2026-10-19

❧

In plain English

  • • FarziPedia turns YouTube videos into illustrated blog posts. The Chrome extension reads the video's transcript and captures frames from your active YouTube tab.
  • • The extension sends the transcript, the video's title and channel, and the frames to Anthropic's Claude API, which writes the post. If you haven't added an API key, they pass through our server at farzipedia.com, which forwards them to Anthropic with FarziPedia's own key and keeps no copy. If you have added your own Anthropic key, they go straight from your browser to Anthropic, and your key is never sent to us.
  • • The finished blog post and its frames are then published on farzipedia.com, in a public library anyone can browse — so only render videos you are happy to see listed publicly.
  • • We do not collect your name, email, account info, location, or YouTube identity. We do not sell user data or share it beyond what this page describes. We do not use cookies, analytics, or trackers in the extension or on this website.
  • • If you want a folio removed from farzipedia.com, email us — see Contact.

1. Who we are

FarziPedia ("we", "us", "the service") is an open-source project that provides:

  • A Chrome browser extension distributed via the Chrome Web Store and at github.com/FarziBuilder/farzipedia-extension.
  • A website and backend service (FastAPI) at farzipedia.com that hosts the public folio library. The same service also answers at farzipedia.onrender.com, its address at our hosting provider.

Source code for both is public at github.com/FarziBuilder. This policy applies to both the extension and the website.

2. What information we collect

FarziPedia is designed to collect the minimum information necessary to do its job. We do not collect personally identifying information from end users.

2.1 Information the extension reads from your browser

When you open the extension's popup on a YouTube video and press Render the Codex, the extension reads:

  • The video's public metadata — its video ID, title, channel name, description, duration, and the URL of the watch page you have open.
  • The video's caption track (transcript), fetched from YouTube's existing endpoints from inside your YouTube tab, using your current browser session.
  • Screenshots of the video element at automatically-chosen timestamps (at most 60 per video), captured via Chrome's tabs.captureVisibleTab API. YouTube's surrounding UI, player controls, and captions are hidden first; each capture is then cropped to the video and saved as a JPEG no larger than 1600 pixels on its longest side.

If you start a playlist run, the extension also downloads the public playlist page from youtube.com (without your YouTube cookies) to read the playlist's title and video list, then loads up to 50 of its videos one after another in your YouTube tab, muted, and reads each one as above.

The extension does not read information from any other tab, never runs without your explicit click on the toolbar icon, and does not access your YouTube account details, browsing history, bookmarks, or any other browser data. The only browser permissions it requests are storage and unlimitedStorage (to keep your optional API key and recent posts), activeTab and scripting (to read and capture the YouTube tab you are on), and access to youtube.com and farzipedia.com (including their subdomains), api.anthropic.com, and farzipedia.onrender.com.

2.2 What goes to Anthropic, and your optional API key

Claude, Anthropic's AI model, writes each post. For each video the extension sends Anthropic the timestamped transcript, the video's title and channel name, and the screenshots described in §2.1, for the claude-opus-4-8 model. How the request gets there depends on whether you have added your own Anthropic API key in the popup:

  • Without your own key (the default). The extension sends the request to our server (farzipedia.com/v1/proxy/messages), which adds FarziPedia's own Anthropic API key, forwards the request to Anthropic, and returns Anthropic's answer. Our server handles the request in memory only: it does not store or log the transcript, the screenshots, or the generated text. Anthropic handles the data as part of FarziPedia's Anthropic account.
  • With your own key. The extension sends the request directly from your browser to Anthropic (api.anthropic.com) with your key. It does not pass through our server, the key is never sent to farzipedia.com, and Anthropic bills the usage to your account and handles the data as part of it.
  • Where your key is kept, if you add one. In Chrome's chrome.storage.sync area, which belongs to your Chrome profile. If Chrome Sync is turned on, Chrome also copies it to your other signed-in Chrome browsers through your Google Account. The Clear button in the popup deletes it.

Either way, Anthropic receives and handles that data under its own terms and privacy policy — see anthropic.com/legal/privacy. If you use your own key, you can revoke it at any time in your Anthropic account's API key settings.

2.3 Information farzipedia.com receives

When Claude has written the post, the extension uploads it to farzipedia.com (falling back to www.farzipedia.com or farzipedia.onrender.com, the same service, if that address does not respond). The upload contains:

  • The generated blog post, plus details about the source video: its ID, URL, title, channel name, and duration, how many frames were captured, when the post was generated, and which extension version and Claude model produced it.
  • The captured frames — the same screenshots that were sent to Anthropic.
  • For a playlist run, the playlist's title, used to name the tome (collection) that run's posts are filed under.
  • Standard HTTP request metadata generated by your browser/network: your IP address, User-Agent string, and the request timestamp. Our code uses the IP address only for rate-limiting (see §5.1) and does not write it to disk.

The upload is what places the post in the public library. If it fails, the extension tells you and keeps the post on your device (§2.4). Simply visiting farzipedia.com sends our server the same standard request metadata; the site sets no cookies.

2.4 Information the extension stores locally on your device

The extension keeps the following in its own storage inside your browser:

  • Your Anthropic API key, if you add one, in chrome.storage.sync (see §2.2).
  • Your most recent generated posts (blog content + captured frames), in chrome.storage.local: the last five, or every post from your latest playlist run plus up to five earlier ones. They are listed on the extension's local library page (result/result.html), so you can reopen them, including any that could not be published. Older posts are deleted automatically.
  • A short record of the most recent run (whether it succeeded, when it finished, how many posts were published, any error, and up to 30 warning messages), which the popup shows after a failed run.
  • While a run is in progress, its status (current step, warnings, and the video's ID, title, channel, description, and duration), so the popup can re-attach if you close and reopen it. This is deleted when the run ends.

Apart from the generation request (§2.2), the upload (§2.3), and Chrome's own syncing of your key, none of this leaves your browser. Removing the extension deletes all of it.

2.5 What we do NOT collect

  • Your name, email, age, gender, address, phone number, or other personally identifying information.
  • Your YouTube account identity, login state, watch history, or subscriptions. A playlist is read only when you paste its link to start a playlist run.
  • Your Anthropic API key, if you add one — the extension keeps it in your Chrome profile and sends it only to Anthropic.
  • Any payment, financial, or health information.
  • Your browsing history, the contents of other tabs, or any data from non-YouTube websites.
  • Geolocation data.
  • Keystrokes, mouse movements, or any form of analytics on user behaviour.

3. How we use the information

We use the data described in §2 solely for the following purposes:

  • Generating your blog post. The transcript, the video's title and channel, and the frames are sent to Anthropic's Claude API so Claude can write the post about the video: through our server with FarziPedia's key, or directly from your browser with your own key (§2.2).
  • Displaying your folio publicly. The resulting blog post and its frames are saved to farzipedia.com and shown in the public library, which any visitor may browse; posts from a playlist run are grouped into a tome named after the playlist. This happens for every post you generate, including posts about unlisted videos.
  • Rate-limiting abuse. Our server keeps the IP address of each upload, and of each request it forwards to Anthropic, in memory (never on disk) to cap how many requests one address can make per minute.
  • Operating the service. Standard request logs (HTTP status, request path, response time, and possibly your IP address) are produced by our hosting provider (Render) and its network (Cloudflare) for diagnosing outages.

We do not use any of this data for advertising, profiling, behavioural targeting, training machine-learning models, or sale to third parties.

4. Third parties

4.1 Service providers we use

FarziPedia relies on the following providers. We do not send them personal information, but some of them see your requests directly, so you should be aware of their independent privacy practices:

  • Anthropic, PBC — receives the transcript, the video's title and channel, and the frames, either forwarded by our server with FarziPedia's key or sent directly from your browser with your own key, and returns a generated blog post. Anthropic handles that data under its own terms and privacy policy (see §2.2). See anthropic.com/legal/privacy.
  • Render, Inc. — hosts the farzipedia.com FastAPI service and the persistent disk on which published folios are stored. See render.com/privacy.
  • Cloudflare, Inc. — Render serves farzipedia.com through Cloudflare's network, so Cloudflare may observe request metadata (IP, User-Agent, requested URL). See cloudflare.com/privacypolicy.
  • Google LLC (Fonts) — the extension's popup and result pages, and this website, link to Google Fonts stylesheets (Cinzel, Cormorant Garamond, IM Fell English, Italianno, and on this website a few more) for typography. When those pages render, your browser requests the font CSS and font files from Google's CDN, which may observe request metadata. See policies.google.com/privacy.
  • Tailwind CSS CDN — this website (not the extension) loads its styling script from cdn.tailwindcss.com, which may observe request metadata.
  • YouTube / Google LLC — the extension reads transcript and frame data from YouTube pages you have opened in your own browser session; transcript requests are made from inside that tab, as YouTube's own player makes them. For a playlist run, the extension also downloads the public playlist page and loads each video in your tab. This website shows video thumbnails served by YouTube (img.youtube.com).

4.2 We do not sell user data

We do not sell, rent, license, or otherwise transfer user data to third parties for marketing, advertising, lending, creditworthiness assessment, or any purpose unrelated to the single purpose described above. We do not have advertising or analytics SDKs of any kind.

5. Data storage and retention

5.1 On our backend

  • Published folios (the blog JSON with its source-video details + captured frames) and tome names are stored on a Render persistent disk attached to the farzipedia.com service. They are kept indefinitely so they remain readable from the public library, unless you request deletion (§6).
  • IP addresses used by the rate limiter live in the server's memory only. Our code never writes them to disk, and they are cleared whenever the service restarts or is redeployed.
  • Request logs produced by Render and Cloudflare are retained per their default retention policies.

5.2 On your device (extension)

The extension's local storage holds at most your last five posts, or after a playlist run, that run's posts plus up to five earlier ones. Older entries are deleted automatically. Your API key, if you added one, stays until you click Clear in the popup. Removing the extension deletes everything it stored on your device.

5.3 Security

The extension connects to Anthropic and farzipedia.com only over HTTPS, and our server forwards requests to Anthropic over HTTPS. If you add your own Anthropic API key, it is kept in Chrome's extension storage and sent only to Anthropic; it never reaches our server. Anyone with access to your Chrome profile could read it, so we recommend setting a monthly spend limit on the key in your Anthropic account. Source code is open for inspection at the GitHub links above.

6. Your rights and how to exercise them

Because we do not collect identifying information, we cannot look up "your" data by name or account — but you can still:

  • Remove a folio from the public library. Email faraaz@elixarsystems.com with the folio's URL (the /blog/<id> link). We will delete it within seven (7) days.
  • Delete what the extension stores on your device. Click Clear in the popup's API key section to delete your key, or remove the extension (chrome://extensions/ → FarziPedia → Remove) to delete your key and all locally stored posts.
  • Revoke your API key. Disable or delete it at any time in your Anthropic account; the extension can then no longer use it.
  • Stop using the service. Uninstalling the extension stops all data collection by FarziPedia immediately. No FarziPedia account or subscription exists to cancel.
  • Request information. If you live in a jurisdiction with data-subject access rights (GDPR, CCPA, etc.), contact faraaz@elixarsystems.com. Because we hold no identifiers, our response is usually limited to confirming which folios match a description you provide.

7. Children's privacy

FarziPedia is not directed at children under 13 (or the equivalent minimum age in your jurisdiction). We do not knowingly collect information from children. If you believe a child has used the service and you would like associated content removed from farzipedia.com, contact us at faraaz@elixarsystems.com.

8. International users

The farzipedia.com backend is hosted in the United States (Render's Oregon region). If you use the service from outside the United States, your published folios and the request data our server receives will be transmitted to the United States, and folios are stored there. Anthropic is a US company and handles the data the extension sends it under its own privacy policy, which describes where that data is processed. We rely on standard HTTPS transport for all such transfers.

9. Changes to this policy

If we materially change how we collect, use, or share information we will update this page and, where reasonable, bump the extension's version number with a note in the release. The "Last updated" date at the top of this page always reflects the current revision. Material changes will be noted in the GitHub commit history of FarziBuilder/farzipedia for transparency.

The 2026-10-19 revision describes extension version 0.5.5, which no longer needs your own API key: without one, it sends generation requests through our server, which forwards them to Anthropic with FarziPedia's key. The 2026-10-17 revision described version 0.5.4, which required your own key and called Anthropic directly. Earlier versions sent the transcript and frames through a proxy on our server at farzi.me, a domain we no longer use.

10. Contact

Privacy questions, removal requests, or anything else you'd like to raise:

❦

Crafted in good faith, for the curious reader.