How to Download YouTube Comments Without an API Key

Every method compared

The Google Cloud Console tax

YouTube's Data API v3 is capable and well documented, but reaching your first comment takes a Google Cloud project, the API enabled on it, credentials generated, and — if the tool will be used by anyone other than you — an OAuth consent screen that may need review before it leaves testing mode.

Then there's quota. Projects get 10,000 units per day shared across endpoints. A commentThreads.list call costs 1 unit, which sounds roomy until you pull reply threads: replies need their own calls, so a channel-wide sweep with replies burns units far faster than the headline number suggests. For a creator who wants to read last week's feedback, that's a lot of ceremony for a small job.

Exporting in four steps

1. 1

   Install YouTube Comments Scraper

   Free on the Chrome Web Store. No Google Cloud project, no OAuth consent screen, no API key.

2. 2

   Open it on any public video

   The side panel auto-fills the video you're watching, or paste any public video URL.

3. 3

   Choose sort and reply depth

   Sort by top or newest, and set depth — 0 for top-level comments only, higher to pull the reply threads underneath.

4. 4

   Export to CSV or JSON

   One click writes the five columns below, with nested replies included when depth is above 0.

What you get back

username,comment,likes,hasReplies,date
@marcusbuilds,"The section at 4:12 finally made subnetting click for me.",1842,true,"2 days ago"
@priya.codes,"Would love a follow-up on the migration path from v2.",417,true,"1 day ago"
@quietcoder99,"Timestamps in the description would help a lot.",96,false,"3 days ago"
@dev_anna,"Second this — the v2 migration is the part everyone gets stuck on.",58,false,"1 day ago"

What creators actually use this for

Automating it with the API

For channel-wide sweeps or scheduled pulls, FetchLayer's YouTube Comments API is a single POST /youtube/comments endpoint taking a video URL, a sort order, a reply depth, and a page count.

curl -X POST https://api.fetchlayer.dev/youtube/comments \
  -H "Authorization: Bearer $FETCHLAYER_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "videoUrl": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
    "sort": "top",
    "depth": 1,
    "pages": 3
  }'

From Node, finding the most-liked comment on a video:

const res = await fetch("https://api.fetchlayer.dev/youtube/comments", {
  method: "POST",
  headers: {
    "Authorization": "Bearer " + process.env.FETCHLAYER_KEY,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    videoUrl: "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
    sort: "top",   // or "newest"
    depth: 1,      // 0 = top-level only
    pages: 3,      // each page is one scroll for more comments
  }),
});

const comments = await res.json();
const mostLiked = comments.sort((a, b) => b.likes - a.likes)[0];
console.log(mostLiked.username, mostLiked.likes, mostLiked.comment);

No Cloud project, no quota units to budget, and you pay for the call rather than the size of the response — a video with 5,000 comments costs the same per page as one with 50.

Reading comments inside an AI agent

Comment analysis is mostly thematic clustering, which is what LLMs are good at. FetchLayer's MCP server exposes the endpoint as a native tool, so Claude, Cursor, or Windsurf can pull a video's comments and summarise them without an export step in between:

{
  "mcpServers": {
    "fetchlayer": {
      "url": "https://mcp.fetchlayer.dev",
      "headers": {
        "Authorization": "Bearer sk-..."
      }
    }
  }
}

Frequently asked questions