MCP
Ghostwrite in your exec's real voice
Feed the exec's actual talk transcripts in, then draft posts that sound like them because they're built from them.
library_import_youtube_videoget_youtube_transcriptchatThe problem
Every ghostwritten exec post has the same tell: it reads like the comms team. The syntax is too clean, the hedges are gone, the pet phrases are missing — and the audience, which includes people who've heard the exec speak, clocks it instantly. The exec knows too: they read the draft, feel the wrongness, and either rewrite it (defeating the purpose) or approve it flatly (defeating the point).
Voice isn't mystical; it's patterns — sentence rhythm, favorite constructions, the specific way this person builds an argument and lands a point. Those patterns are documented at length in every talk and interview the exec has given. Ghostwriting that starts from the transcripts starts from the voice.
How veridive MCP helps
library_import_youtube_video and get_youtube_transcript build the voice corpus: the exec's talks, interviews and panels, imported once, transcripts pulled — hours of documented voice, in text.
chat does the drafting against it: with the corpus in the conversation, drafts inherit the exec's actual constructions — the rhythm, the hedges, the way they set up a contrast — and, crucially, their positions: what they've actually said about the topic, so the post extends their thinking rather than inventing it. Review cycles shrink from rewrites to touches, because the draft already sounds like the person whose name goes on it. That's the whole job.
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.
- 2Build the voice corpus
The exec's substantive appearances, imported; transcripts pulled — done once, reused always.
Import these talks by <exec> into my library, then pull their transcripts: <video URLs> — use veridive's library_import_youtube_video. - 3Extract the voice patterns
The stylistic fingerprint, made explicit — usable across every future draft.
From these transcripts: how does <exec> actually talk? Sentence rhythm, recurring constructions, pet phrases, how they open and land arguments. - 4Draft grounded in voice and positions
Posts that sound like them and extend what they've actually said.
Draft a LinkedIn post on <topic> in <exec>'s voice — using their patterns and consistent with the positions they took in these transcripts.
Problems it solves
- Ghostwritten content reads as comms-team output and audiences clock it.
- Execs rewrite drafts that don't sound like them — the process defeats itself.
- Voice lives in the writer's impression of the exec instead of the exec's documented speech.
- Drafts invent positions the exec never took, and legal review becomes the last line of defense.
Why it's valuable
- Drafts inherit documented voice patterns — the wrongness feeling disappears from review.
- Positions come from what the exec actually said; the post extends, never invents.
- The voice corpus is a standing asset: every future draft starts from it.
- Exec trust in the process compounds — which is what makes a ghostwriting program sustainable.
Example prompts
Import <exec>'s talks and pull transcripts: <URLs> — use veridive's library_import_youtube_video.Extract the voice fingerprint: rhythm, constructions, argument patterns.Draft a post on <topic> in that voice, consistent with their stated positions.FAQ
Which tools build the corpus?
library_import_youtube_video to pin the appearances, get_youtube_transcript for the text (free once imported), then chat for the drafting — the writing half needs no retrieval, just the corpus in context.
How much material makes a workable voice corpus?
Three or four substantive talks is enough for the fingerprint; more sharpens it. Prefer unscripted formats — panels and podcasts carry more voice than rehearsed keynotes.
Does this pair with the spokesperson-drift analysis?
Same corpus, two directions: drift analysis audits what the exec says against the messaging doc; this writes new content grounded in how they say things. Comms teams end up running both on one archive.
Related use cases
Import their appearances, extract key quotes, and check delivered-message vs intended-message drift.
Sweep your executive's appearances, archive them, and keep the quotes-on-record list current.
Authors give away the thesis on the promo circuit; find their talks and summarize — closer to the source than any summary app.
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.