> ## Documentation Index
> Fetch the complete documentation index at: https://linkly.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Using the Browser Extension

> Linkly AI Clipper: turn any web page into cleanly formatted Markdown in one click — preview, copy, and download it in the side panel, or send it straight into your Linkly AI Space.

## What It Does

**Linkly AI Clipper** is Linkly AI's browser extension, and it does exactly one thing: **save web pages as agent-friendly Markdown**.

One click turns the page you're on into cleanly formatted Markdown that you can preview, copy, or download from the side panel. It's more than just saving a page — article text, images, translations, video subtitles, even your conversations with AI all become a single clean Markdown file an agent can work with straight away.

<CardGroup cols={2}>
  <Card title="Saved in a Format Agents Can Read" icon="file-code" iconType="duotone">
    YAML front matter and a clear section structure, with dedicated handling for sites like X and Weibo
  </Card>

  <Card title="Straight into Your Library" icon="folder-tree" iconType="duotone">
    With the desktop app installed, clips land right in your Linkly AI Space and get indexed alongside everything else on your computer
  </Card>

  <Card title="Images Come Along" icon="image" iconType="duotone">
    Pictures from the article are downloaded into the same folder, so the piece still reads properly offline
  </Card>

  <Card title="Everything Happens Locally" icon="shield-check" iconType="duotone">
    Extraction, translation, and outline generation all run inside your browser — nothing goes through our servers
  </Card>
</CardGroup>

***

## Installing

Install it from the Chrome Web Store: [**Linkly AI Clipper**](https://chromewebstore.google.com/detail/linkly-ai-clipper/akplbgdjkgkakgoemhpjpphfilipdbda).

| Item                      | Requirement                                                                                           |
| ------------------------- | ----------------------------------------------------------------------------------------------------- |
| **Browser**               | Chrome 116 or later (the side panel API needs it)                                                     |
| **On-device AI features** | On-device translation and the AI outline need a recent Chrome (the built-in models must be available) |
| **Other browsers**        | Chromium-based browsers such as Edge and Brave can install it straight from the Chrome Web Store      |
| **Sign-in / payment**     | Neither is required — see the [FAQ](#faq)                                                             |

<Warning>**Firefox isn't supported yet.** The extension is published on the Chrome Web Store only; there's no Firefox build, so don't go looking for it in Firefox Add-ons.</Warning>

***

## Clipping Your First Page

<Steps>
  <Step title="Open an article page">
    Any article, tweet, forum answer, or YouTube video page will do.
  </Step>

  <Step title="Click the L icon in the toolbar">
    Click the extension icon in the top-right corner of the browser. The side panel opens on the right and immediately starts extracting the body of the current page and converting it to Markdown.

    <Note>
      **There's no context menu and no keyboard shortcut.** Clicking the toolbar icon is the only way to open the side panel, so it's worth pinning the extension to your toolbar (click the pin in the extensions menu).
    </Note>
  </Step>

  <Step title="Check the result, then copy or save">
    The default view is **Preview** — a properly typeset reading view. The toolbar also offers **Edit source** (switch to raw Markdown and edit it directly), **Metadata** (source URL / site / author / publication date), **Table of contents**, and four text-size steps.

    Once it looks right, click **Copy** to put the whole Markdown on your clipboard and paste it straight to an AI assistant, or click **Download** to save it as a file — where it ends up is covered in [the next section](#where-clips-are-saved).
  </Step>
</Steps>

<Note>**Switching tabs re-converts automatically.** If you'd rather it didn't, turn off **Settings → Auto-refresh**; after that you click **Re-convert** yourself.</Note>

<Warning>**Selection clipping isn't supported.** The extension always extracts the full body of the page — there's no option to clip only the part you highlighted. To keep just a portion, switch to **Edit source** in the side panel and delete the rest yourself.</Warning>

***

## Where Clips Are Saved

This is the most confusing part: **the same Download button takes two completely different paths depending on whether the desktop app is running.**

| Situation                                             | Where it's saved                              | What follows                                                 |
| ----------------------------------------------------- | --------------------------------------------- | ------------------------------------------------------------ |
| **The desktop app is running**                        | `~/LinklyAI/Clips/<clip title>/`              | Indexed automatically, then searchable and retrievable by AI |
| **The desktop app isn't running, or the save failed** | `~/Downloads/Linkly/` (your downloads folder) | Just ordinary files — **not** indexed                        |

Each clip is **a folder**, containing the `.md` file and its images.

### How to Tell Which Path It Took

A toast appears at the bottom of the side panel after saving, one of four:

| Toast                                                        | What it means                                            |
| ------------------------------------------------------------ | -------------------------------------------------------- |
| Saved to Linkly Space                                        | It made it into your library                             |
| Client offline — saved to browser downloads                  | The desktop app wasn't detected, so it fell back         |
| Couldn't save to Linkly desktop — saved to browser downloads | It was detected but the transfer failed, so it fell back |
| Saved locally                                                | You turned **Save to Linkly Space** off in settings      |

You can also check proactively: open **Settings → Download settings**, and the **Save to Linkly Space** row shows the current state — **Online** plus the desktop app's version number means it's connected, **App not detected** means it isn't. Next to it sit two buttons, **Re-check** and **Launch desktop app**.

<Warning>**The fallback is one-shot; nothing gets retried.** A clip that went to your browser downloads folder doesn't join a queue, and it won't be swept into your library later when you do start the desktop app. To get it indexed, either move the files into your Linkly AI Space yourself, or add the downloads folder itself as an [indexed folder](/docs/en/indexing).</Warning>

<Note>The fallback **loses no content** — the Markdown and the images are all there, just somewhere else. The subfolder name inside your browser downloads folder (`Linkly` by default) can be changed under **Settings → Download settings → Default folder**. The `Clips/` directory on the library side **can't be configured separately**; it follows wherever your Linkly AI Space is.</Note>

Once a save succeeds, an **Open save location** button appears in the toolbar; click it and this clip's folder opens in your system file manager (older desktop app versions don't get this button).

<Warning>The desktop app has **no** interface for browsing clips. To see what you've clipped, open `~/LinklyAI/Clips` in your file manager. The clipped content itself is of course searchable from the [launcher](/docs/en/use-launcher) and in [Chat](/docs/en/use-chatbot).</Warning>

***

## What a Clip File Looks Like

The top of the Markdown carries a block of YAML front matter declaring compatibility with Open Knowledge Format v0.1 — metadata written for AI agents, so they can tell at a glance what the file is and where it came from:

```yaml theme={null}
---
type: "Video Transcript"
title: "What is Linkly AI?"
resource: "https://www.youtube.com/watch?v=HnGMkWaWvVo"
timestamp: 2026-07-28T12:11:16Z
site: "YouTube"
author: "Linkly AI"
---
```

The full field list: `type` (content type), `title`, `description` (summary, omitted when it duplicates the title), `resource` (source URL), `tags`, `timestamp` (when it was clipped), `site`, `author`, `published` (the original publication date), and `lang`. A field is only written when it has a value; the rest are simply left out.

`type` is decided by the site. Common values are `Web Article`, `WeChat Article`, `X Thread`, `Zhihu Article`, `Video Transcript`, and `AI Conversation`.

<Tip>Don't want front matter? Turn off **Settings → Export content → Include frontmatter** and the export is plain body text.</Tip>

### Images

Images in the article body are **downloaded locally**, and the references in the Markdown are rewritten to local relative paths — no remote links are kept, so the images still display offline.

* The images sit **flat in the same folder** as the `.md`; there's no `image/` subfolder
* The naming rule is `<title>-visual-N.<ext>`
* A single clip holds at most 200 files
* If you only want the text, turn off **Settings → Export content → Include images**: no image files are downloaded, and the image references are stripped out of the Markdown as well

***

## Site-Specific Handling

Generic body extraction (stripping navigation, ads, and recommendation blocks) is enough for most sites, but some have unusual structures and get dedicated handling:

| Site                         | What's handled                                               |
| ---------------------------- | ------------------------------------------------------------ |
| **ChatGPT / Claude**         | Whole-conversation export, or a single reply                 |
| **X (Twitter)**              | Full thread capture, with long-form posts handled separately |
| **YouTube / Bilibili**       | Subtitles transcribed into timestamped Markdown              |
| **WeChat Official Accounts** | Body text, author, publication date                          |
| **Zhihu**                    | Answers and column articles                                  |
| **Lark Docs**                | Read from the document data model rather than the DOM        |
| **Xiaohongshu**              | Note body plus the image carousel (avatars excluded)         |
| **Weibo**                    | A single post plus its images                                |
| **V2EX**                     | Topic body                                                   |

Code blocks get their language detected automatically and tagged. Every other site goes through the generic extraction path.

***

## Other Features

### On-Device Translation

**Translate** in the side panel toolbar renders the current Markdown into English, Simplified Chinese, Traditional Chinese, Japanese, or Korean, and you can switch back to the original at any time.

It uses **Chrome's built-in translation model** — no API key to configure, and the content never leaves this computer. The first time you use a given language pair the model is downloaded, with progress shown in the side panel. A recent version of Chrome is required.

### Using It as a Reader

The side panel preview is a clean reader in its own right: automatic typography, four text-size steps, dark mode that follows the system, and side-by-side translation. You don't have to save anything — treat it as an ad-free reader if you like.

### AI-Generated Outline

Some pages have no proper heading hierarchy (a wall of plain paragraphs, say), which leaves the side panel's table of contents empty. When that happens, click **Generate AI outline** and Chrome's built-in on-device model infers the section titles and builds a clickable table of contents.

Copy and Download include those AI headings by default; to keep the export strictly faithful to the original, turn off **Settings → Export content → Include AI-generated article headings**. This also needs a recent version of Chrome, and works only on desktop Chrome / Edge.

### Exporting a Single AI Reply

On **ChatGPT** and **Claude** conversation pages, the extension injects a download button next to every AI reply — no need to clip the whole page, one click saves that single reply as Markdown.

The whole conversation can be clipped too (click the toolbar icon and follow the normal flow); its `type` will be `AI Conversation`.

***

## Working with the Desktop App

### Version Requirements

Saving clips straight into your Linkly AI Space needs **a reasonably recent desktop app** — older versions have no endpoint for receiving clips, so the extension treats them as not running and the clip lands in your browser downloads folder. Leaving the desktop app on auto-update is enough; there's no version number to compare by hand.

### How They Talk to Each Other

The extension and the desktop app communicate over **plain local loopback HTTP** (`127.0.0.1`) and **never go through the cloud**. The desktop app applies two admission checks: only direct local connections are allowed, and the extension's origin must be on an allowlist — no other program or web page can connect.

<Note>The port isn't guaranteed to be 60606: that's the default, and it moves up when the port is taken. The extension only probes the 10 ports from 60606 to 60615, so in the unlikely event all of them are held by other programs, the desktop app is judged not running even while it is, and the clip lands in your browser downloads folder.</Note>

### How Soon Can You Search a Clip

Clips saved into your library are indexed automatically, but **not instantly** — the file watcher debounces and indexing queues up, so it usually takes anywhere from a few seconds to a minute. Searching the moment you finish clipping may turn up nothing; wait a little and try again.

The layered indexing mechanism and how to troubleshoot it are covered in [Index Settings](/docs/en/indexing).

***

## FAQ

<AccordionGroup>
  <Accordion title="Do I need to sign in? Does it cost anything?">
    **No sign-in, no payment.** Extraction, translation, copying, downloading, saving into your library — none of it depends on being signed in.

    There is a **Sign in with Linkly AI** entry in the side panel, but at the moment all it does is show your details on the account button; signing in or not makes no difference to any feature.
  </Accordion>

  <Accordion title="Do I have to install the desktop app? What happens to clips when it isn't running?">
    You don't. Extracting the body, translating, copying, and downloading all work without it, so the extension stands entirely on its own. What the desktop app adds is that clips go straight into your library and get indexed alongside the rest of your files, which makes them retrievable by any AI agent afterwards — well worth pairing the two.

    When the desktop app isn't running, clips fall back to your browser downloads folder at `~/Downloads/Linkly/`. **Nothing is lost**, and the side panel says so plainly: **Client offline — saved to browser downloads**. But those files **won't** be swept into your library later when you start the desktop app; if you want them there, move them into your Linkly AI Space yourself.
  </Accordion>

  <Accordion title="I can't find something I clipped">
    Look in this order:

    1. **Your library**: `~/LinklyAI/Clips/<clip title>/` — where clips land while the desktop app is running
    2. **Your browser downloads folder**: `~/Downloads/Linkly/` — the fallback when the desktop app isn't running

    The desktop app has **no** interface for browsing clips; your file manager is the only place to look. Not sure which path a clip took? Check whether the toast at save time read **Saved to Linkly Space** or **saved to browser downloads**.
  </Accordion>

  <Accordion title="I clipped something and searched right away, but nothing came up">
    That's normal. A clip that enters your library takes anywhere from a few seconds to a minute to finish indexing — the file watcher debounces, and indexing tasks queue up. Wait a moment and search again.

    If it still doesn't show up after a long wait, open **Settings → Index** and check whether indexing is stuck; troubleshooting steps are in [Index Settings](/docs/en/indexing).
  </Accordion>

  <Accordion title="Is Firefox supported?">No. The extension is currently published on the Chrome Web Store only. Chromium-based browsers such as Edge and Brave can install it straight from there; Firefox has no available build.</Accordion>

  <Accordion title="Are there keyboard shortcuts or a context menu?">
    Neither. The extension declares no keyboard shortcuts and registers no context-menu items — clicking the extension icon in the toolbar is the only way to open the side panel.

    The one exception is ChatGPT and Claude pages, where a download button is injected next to every AI reply.
  </Accordion>

  <Accordion title="Do the pages I clip get uploaded to your servers?">
    No. Body extraction, translation, and outline generation **all happen inside your browser**.

    Saving has only two destinations: your local Linkly AI Space, or your browser downloads folder. Communication with the desktop app runs over local loopback (`127.0.0.1`) and touches no network service.
  </Accordion>

  <Accordion title="Why does it need permission to read data on all websites?">
    Because you might clip from any site, so the extension has to be able to run its extraction script on the page you're on. It reads the current tab's content **only when you click**, and never scans your browsing in the background.
  </Accordion>
</AccordionGroup>

***

## Further reading

* [Index Settings](/docs/en/indexing) — how clips get indexed once they land in your library
* [Using the Launcher](/docs/en/use-launcher) — the fastest way to find clipped content
* [Using Libraries](/docs/en/use-libraries) — put clips into a specific library so AI can search them by scope
* [Quick Start](/docs/en/quickstart) — installing and setting up the desktop app
* [Linkly AI Space](/docs/en/linkly-space) — the folder clips land in, and how to move it
* [Using the Chatbot](/docs/en/use-chatbot) — have the AI read the pages you clipped
