MCP

AI-agent format

Mine comments for your next ten videos

Comment summarizer surfaces the questions your audience keeps asking under your videos and your rivals' — answered demand, pre-validated.

CreatorContent Strategy & Creatorsrun_youtube_tool
The use case

The problem

Under every decent video sits a focus group that paid for itself: hundreds of comments where your audience tells you, unprompted, what confused them, what they want next, and which thirty seconds they'd watch a full video about. Creators know this in theory. In practice, comment sections get skimmed for praise and trolls, and the request layer — "please make one about X", "you lost me at the part where...", "does this work for Y?" — scrolls past unmined.

Multiply by your whole channel, then by your competitors' channels (whose audiences are your audience), and there's a content calendar sitting in plain sight: ideas pre-validated by the exact people who'd watch them, expressed in the exact words they'd search with.

The mechanics

How veridive MCP helps

run_youtube_tool's comment-summarizer reads the whole section so you don't: recurring questions, requests and confusion points, aggregated and ranked by frequency instead of by recency and rage. comment-sentiment adds the temperature — which themes come with enthusiasm (make that video) versus frustration (fix that gap).

Run it across your recent uploads and the mine produces your backlog; run it across your competitors' biggest videos and it produces theirs — the requests their audience made that they haven't fulfilled. Both passes work on any public video from its URL. The output isn't inspiration; it's demand, counted, in the audience's own phrasing — which doubles as your title draft.

Step by step

How to run it

  1. 1
    Connect veridive MCP

    One-time setup: add the veridive MCP server to Claude, ChatGPT, Cursor or any MCP client and generate your API key. Takes about two minutes.

  2. 2
    Mine your own sections

    Recent uploads, comment summaries, requests extracted and ranked by recurrence.

    Summarize the comments on these videos of mine and pull out every question and video request, ranked by how often it recurs: <video URLs> — use veridive's run_youtube_tool.
  3. 3
    Mine the competition's sections

    Your rivals' audiences are telling them what to make — and they're mostly not listening. Same pass, their biggest videos.

    Same analysis on these competitor videos: what does their audience keep asking for that they haven't made? <video URLs>
  4. 4
    Turn requests into the backlog

    Merge, dedupe, rank — and keep the audience's phrasing; it's your best title language.

    Merge both passes into a ranked idea backlog: request, frequency, source videos, and a working title using the audience's own wording.
Before vs after

Problems it solves

  • Comment sections get skimmed for vibes while the request layer scrolls past unread.
  • Recurring questions are invisible when you read comments video-by-video — recurrence only shows in aggregate.
  • Competitor audiences broadcast unmet demand daily, to nobody.
  • Ideation runs on inspiration when a counted, pre-validated backlog is sitting under every upload.
The payoff

Why it's valuable

  • Every idea arrives pre-validated by the people who'd watch it — demand before production.
  • The audience's phrasing comes with it: titles and thumbnails in the language viewers actually use.
  • Competitor mining finds requests with proven audiences and no supplier.
  • Ten minutes per batch of videos — the focus group runs on a prompt.
Copy & paste

Example prompts

Summarize comments on my videos and rank every recurring request: <URLs> — use veridive's run_youtube_tool.
What does this competitor's audience keep asking for that they haven't made? <URLs>
Merge into a ranked backlog with working titles in the audience's phrasing.
Questions

FAQ

Does this need any import or setup per video?

No — comment-summarizer and comment-sentiment run on any public video straight from its URL. Batch as many as you like in one conversation.

How does it handle huge comment sections?

That's the point — aggregation is what humans can't do at volume. The summarizer works over the section and surfaces recurrence; you read the ranked output, not the raw thousands.

Is mining competitor comments fair game?

It's public audience feedback on public videos — the same signal their team could read and mostly doesn't. Fulfilling unmet requests isn't taking anything; it's answering people faster.

Keep going

Related use cases

Run this in your own assistant

Connect the veridive MCP server to Claude, ChatGPT or Cursor and paste the prompts above — setup takes about two minutes.