MCP
Build a swipe file of the best hooks in your niche
Pull the openers of breakout videos into a transcript swipe file and study them line by line.
deep_searchlibrary_import_youtube_videoget_youtube_transcriptThe problem
Copywriters have kept swipe files for a century: a private collection of ads that worked, studied line by line until the mechanics become instinct. Video creators need the same file for hooks — the first thirty seconds that decide everything — and almost nobody keeps one, because "studying an opener" means scrubbing videos and transcribing by hand.
So hook craft gets absorbed by osmosis: you watch a lot, you get a feel, you can't articulate why one cold open holds and another bleeds viewers. A swipe file changes that — the actual words of your niche's best openers, on paper, comparable. Written down, hooks stop being magic and start being moves: the question tease, the stakes-first open, the mid-action drop-in. Moves can be learned.
How veridive MCP helps
deep_search finds the file's contents: the breakout videos in your niche whose openers earned study. library_import_youtube_video pins them into your library, and get_youtube_transcript delivers the raw material — timestamped segments, so isolating the first thirty seconds of each is trivial, and pulls are free once imported.
Then the study begins, with your assistant as reading partner: lay the openers side by side, name each one's move, notice what your niche's hooks share and where the outliers deviate. The file compounds — every breakout you add sharpens the pattern — and before your next script, twenty proven openers are sitting there to write against. Not to copy: to count the beats the way copywriters always have.
How to run it
- 1Connect 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.
- 2Find the openers worth stealing from
Breakouts and consistent performers in your niche — the videos whose first thirty seconds demonstrably held.
Find the breakout and consistently strong videos in the <niche> space, recent first — use veridive's deep_search. - 3Import and extract the openers
Pin the shortlist, pull the transcripts, isolate the first thirty seconds of each — the swipe file's raw pages.
Import these into my library, then pull each one's transcript and give me just the first 30 seconds of each, verbatim with timestamps: <video URLs> - 4Study the moves
Side by side, named and compared: what move does each opener make, and what do the winners share?
Lay these openers side by side and name each one's move — question tease, stakes-first, cold action, contrarian claim. What patterns do the strongest share?
Problems it solves
- Hook craft gets learned by osmosis — slow, unarticulated, impossible to teach to your editor.
- Studying openers manually means scrubbing and hand-transcribing, so nobody does it.
- Watching hooks isn't reading them: on paper, the mechanics show; on screen, the charisma hides them.
- Every script starts from instinct alone when twenty proven openers could be on the desk.
Why it's valuable
- The copywriter's oldest tool, finally practical for video — a compounding private asset.
- Verbatim text with timestamps: hooks as studyable writing, not as vibes.
- Named moves are teachable — your future editor and collaborators inherit the file.
- Script sessions start against evidence: which of the file's moves fits this video's promise?
Example prompts
Breakout videos in <niche>, recent first — use veridive's deep_search.Import these, pull transcripts, give me each one's first 30 seconds verbatim: <URLs>Name each opener's move and the patterns the strongest share.FAQ
Why import instead of just watching the openers?
The file is the point: imported videos give you free transcript pulls, permanent text you can re-read, compare and annotate. Watching gives impressions; the swipe file gives mechanics.
Isn't studying hooks just learning to clickbait?
Opposite ends of the same skill: clickbait is a promise the video breaks; a great hook is a promise it keeps, made irresistibly. Studying how winners *frame* honest promises is craft, not deception.
How big should the file get?
Start with ten to fifteen openers — enough for patterns to show. Add each new breakout in your niche as it happens; prune ones whose moves you've fully absorbed.
Related use cases
Study how top performers in your niche title and tag, then generate metadata drafts tuned to the same patterns.
Most-replayed segments plus comment sentiment show what actually hooked viewers — pattern-match it before your next upload.
Blog post, X thread, LinkedIn post, newsletter section, Shorts script — one upload, six assets, same day.
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.