MCP
Hunt sources with timestamped quotes
Find who said it, pull the transcript, quote to the second — attribution your editor can verify in one click.
deep_searchlibrary_import_youtube_videoget_youtube_transcriptThe problem
The quote is the story's load-bearing wall. And the era's strange bargain is that more people say more things on the record than ever — hours of panels, podcasts and interviews daily — while finding who actually said the thing has never been harder. You half-remember a founder admitting something remarkable on some podcast last spring. Which founder? Which podcast? Minute what?
Deadline journalism can't afford the archaeology, so the quote gets dropped, or worse, sourced from an aggregator that got it slightly wrong. The primary tape exists, public and timestamped. What's missing is the path from "someone said something like this" to the exact second they said it — fast enough to survive an edit cycle.
How veridive MCP helps
deep_search is the needle-finder: it searches what was said, not video titles. Describe the statement you're hunting — even approximately — and results come back with evidence snippets pinned to timestamps: the moment in each video where the topic is actually addressed. "Someone from the industry admitted X" becomes three candidate clips with names, in minutes.
library_import_youtube_video plus get_youtube_transcript turn the find into copy-ready attribution: import the source, pull the verbatim transcript, and quote the exact sentence with its full surrounding context — the qualifier they attached, the question that prompted it. Your editor gets a link and a timestamp instead of "I'm pretty sure they said this," and the correction desk never hears about you.
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.
- 2Hunt the statement
Describe what you're looking for the way you remember it — the transcript-level search finds it from meaning, not from titles.
Find video where <person or 'an executive at X'> discusses <the thing you half-remember> — point me at the exact moments — use veridive's deep_search. - 3Verify at the source
Import the candidate video and pull the verbatim passage with context — the sentence, its qualifier, the question it answered.
Import this into my library and get the transcript — show me the full passage around the <topic> statement, timestamped: <video link> - 4File with bulletproof attribution
Quote, timestamp, link. The fact-check takes your editor one click instead of one argument.
Format the quote for publication: verbatim sentence, speaker, source video and timestamp.
Problems it solves
- The half-remembered quote — you know it was said, but which person, which show, which minute?
- Title search is useless for statements: the remarkable sentence is never in the video's title.
- Aggregator paraphrases drift; quoting them inherits their errors under your byline.
- Deadline pressure kills the archaeology, so strong quotes die unfound or unverified.
Why it's valuable
- Meaning-level search over the spoken record — the needle-in-haystack problem, mechanized.
- Attribution ships with a timestamp and link: verification is one click, not one debate.
- Full context travels with the quote — the qualifier and the prompting question survive.
- Minutes per hunt, which means strong quotes make it into deadline copy instead of the cutting-room floor.
Example prompts
Find where <person> discusses <statement> — exact moments — use veridive's deep_search.Import <video link>, pull the transcript passage around the <topic> statement, timestamped.Format for publication: verbatim quote, speaker, source, timestamp.FAQ
What if I don't know who said it?
Describe the statement and whatever context you have — the search works at the meaning level across transcripts, so 'a retail CEO admitting online returns lose money' finds candidates even without a name.
Is the transcript reliable enough to quote from?
Treat it as the finding aid, then confirm the exact wording against the video at the timestamp — that final listen is journalism. The workflow's job is getting you to the right second with the draft text in hand.
Which tools does this chain?
deep_search to locate the moment (evidence timestamps), then library_import_youtube_video + get_youtube_transcript for the verbatim passage — transcript pulls are free once the video is imported.
Related use cases
Follow the citation chain to the original spoken statement — publish the correction with the receipt.
Skip the paraphrase cycle: pull the primary transcript and quote the exact sentence, timestamped.
Every appearance, the stories they always tell, the questions they dodge — ask what no one has.
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.