Reverse-engineer any web image into a ready-to-use AI art prompt.
Last updated: 2026-09-15
Promptary is a Chrome extension that reverse-engineers images into AI art prompts and helps you organise them. This policy explains exactly what happens to your data.
The short version:
Promptary can reach a model in two ways, and the difference matters for your privacy. You choose in Settings → Model service.
| Your own API key | Built-in free channel | |
|---|---|---|
| Who runs the model service | You choose (DeepSeek, OpenAI, OpenRouter, self-hosted…) | We do, using a DeepSeek model |
| Where your image goes | Directly from your browser to that service | Through a relay server we operate |
| What we can see | Nothing — the request never touches us | The request passes through, but is not retained |
| Do you need an account | No | No — an anonymous device ID is generated locally |
| Daily limit | None | Yes (a few images per day) |
The built-in channel exists so you can try Promptary without signing up for a model service first. If you would rather nothing pass through us, use your own API key — that path is fully direct, exactly as it always was.
Regardless of which channel you use, we do not collect:
There is no tracking of any kind.
Everything Promptary saves stays on your own computer:
| Data | Where it is stored | Notes |
|---|---|---|
| Saved images, prompts, tags, categories | Your browser’s local IndexedDB | Never uploaded anywhere |
| Your API key and model settings | chrome.storage.local |
Stored separately from your library |
| Interface preferences | chrome.storage.local |
Theme, language, etc. |
Your API key is never included in exported backups, because settings and library data are stored separately by design.
Deleting items moves them to an in-extension Trash; emptying the Trash permanently removes them. Uninstalling the extension deletes all of the above, since it all lives inside the extension’s own storage.
When you ask Promptary to analyse an image, that image is sent to the AI model service that you configured in Settings.
Promptary supports any OpenAI-compatible API. You supply the endpoint (baseUrl), the API key, and the model name.
Analysing an image sends it through a relay server we operate, which forwards it to DeepSeek and returns the result. See the next section for exactly what that server does and does not record.
By default, Promptary first gives the model service the image’s public URL, letting the service fetch it directly. If that fails (for example, the site blocks hotlinking), Promptary downloads the image locally and uploads it as base64 instead. You can force either behaviour in Settings.
Note that with URL pass-through, the image itself is never transmitted by us at all — only its address is. The bytes are fetched by the model service directly from the original website.
This section applies only if you use the free channel. If you use your own API key, none of it applies to you.
What the relay server does
It checks your daily quota, forwards your request to DeepSeek, and passes the answer back. That is its entire job.
What it records
| Recorded | Why | Kept for |
|---|---|---|
| An anonymous device ID (a random UUID generated by your browser) | To count how many images you have analysed today | 7 days, then deleted automatically |
| The number of requests made that day | Daily quota | 7 days |
The device ID is random, is not derived from anything about you, and is not linked to any account — there are no accounts. Reinstalling the extension generates a new one.
What it does not record
Security and retention
Requests are sent over HTTPS. The only data that persists is the quota table described above, which is deleted on a rolling 7-day basis.
Chrome shows you a permission warning for this extension. Here is what each one is actually for:
| Permission | Why Promptary needs it |
|---|---|
contextMenus |
Adds the right-click menu items (“Reverse-engineer this image”, “Save this image”, and so on). This is the main way you use the extension. |
storage |
Saves your settings and API key locally. |
sidePanel |
Hosts the main interface in the browser’s side panel. |
activeTab |
Reads the current tab’s URL and title when you trigger an action, so saved items can record where they came from. Only used at your request. |
scripting |
If you trigger reverse-engineering from the right-click menu on a page that was already open before the extension was installed, the content script is not yet present. This permission lets Promptary inject it once so the action still works. |
<all_urls> |
See below. |
<all_urls> is requiredPromptary works on any website, because you might find a useful reference image anywhere — an art site, a social feed, a blog, a shop.
Concretely, this permission allows Promptary to:
It does not allow Promptary to do anything in the background. Promptary never reads pages you have not explicitly acted on, and never sends page content anywhere on its own. All of the above happens only in direct response to your clicks.
For the built-in channel, the only outbound destination is our relay server. With your own API key, the only outbound destination is the endpoint you configured yourself.
Promptary integrates with no third-party analytics, advertising, or tracking services.
Two third parties may be involved in analysing an image, depending on your channel:
We do not share, sell, rent, or trade your data. With your own API key we never receive it at all. With the built-in channel we relay it without retaining it, and use it for no purpose other than producing the answer you asked for.
Promptary is not directed at children and collects no personal information from anyone, including children under 13.
If this policy changes, the “Last updated” date above will change, and the revision will be visible in the project’s public repository.
Promptary is an open-source project. Its source code is public, so every claim in this document can be verified directly — including the relay server’s source, which is in the server/ directory of the repository.