ClawLink
TL;DR
| Question | Answer |
|---|---|
| What does it do? | Your agent reads the comments under a YouTube video and groups the requests and questions viewers repeat. |
| Who is it for? | YouTubers planning content, and product teams reading reactions to a review of their product. |
| What do I connect? | Nothing. It reads public data with ClawLink's own provider. |
| What does it cost? | About $0.004 of ClawLink credit per run. New accounts get $0.50 free. |
| Does it work? | Yes. We ran it on 2026-10-03; the results are below. |
| Which agents? | OpenClaw, Hermes, Claude, Codex, ChatGPT and any MCP client connected to ClawLink. |
How do you try it?
- Set up ClawLink once. Add it to your agent from the install page. This job needs no connected app.
- Send this prompt to your agent. Click the card to copy it.
It works the same in OpenClaw, Hermes, Claude and Codex.
How does the agent do it?
The agent makes 2 calls through ClawLink. You do not pick the tools: the agent finds them by searching ClawLink, reads their input schema and runs them.
| Step | Tool | What it does | Price |
|---|---|---|---|
| 1 | youtubedata_channel_videos | Finds the latest video when you name a channel instead of a URL. | $0.002 |
| 2 | youtubedata_video_comments | Returns a page of top comments with likes and reply counts. | $0.002 |
What happened when we ran it?
We sent this job through ClawLink on 2026-10-03. These are the real calls and what came back, not an example.
| Call | Input and result | Time | Cost |
|---|---|---|---|
youtubedata_video_comments | MKBHD's Xiaomi 18 Fold video, 20 top comments | 2.8 s | $0.002 |
What the data showed:
- About 8 of the 20 top comments were about Xiaomi copying the iPhone Duo design.
- The most liked comment (16,000 likes, 135 replies) joked about YouTube's new A/B test of video versions.
- There were no clear requests for future videos: the agent should report that rather than make some up.
How do you write the prompt?
- Ask for requests, not a summary. "What do viewers ask for" pulls out the useful comments.
- Use likes as votes. Ask the agent to weigh each request by the likes its comments got.
- Several pages for big videos. One page is a sample; ask for three pages on a video with thousands of comments.
What can go wrong?
- Top comments are not all comments. The first page is the most liked, which favors jokes. Ask for newest too.
- Comments turned off. Some videos have comments disabled; the agent should say so.
- Some videos get reactions, not requests. In our run, the comments on a phone review were jokes and opinions about copying, with no feature requests. The agent should say so instead of inventing requests.
- Near-duplicates. The same joke appeared twice in one page. Ask the agent to merge repeats before counting.
How does this compare with other ways to do it?
| Option | How it works | Cost | Best for |
|---|---|---|---|
| ClawLink + your agent | One prompt, grouped requests | $0.002 per page | Reading any public video |
| YouTube Studio | Comments on your own videos | Free | Your own channel |
| Read them yourself | Scroll | Free | Small videos |
Frequently asked questions
Do I need a YouTube account?
No. This job reads public YouTube data through ClawLink's own data provider, so there is nothing to connect. It runs on your ClawLink credit; every new account gets $0.50 of free credit on its first data call.
Can it reply to comments?
Not with this tool. Replying needs your YouTube account connected.
Does it read replies?
Yes, there is a separate tool for comment replies.
Does it work on Shorts?
Yes.
Can it compare comments across videos?
Yes, ask for several videos; each page is one call.