MCP

AI-agent format

Turn recorded all-hands into meeting notes

Import the recording, generate meeting notes with decisions and action items, archive searchable — done before lunch.

Ops leadTeam Knowledge & Library Opslibrary_import_youtube_videorun_youtube_tool
The use case

The problem

The all-hands happened, the recording exists, and now the epistemics get weird: half the company watched live and remembers different emphases; the other half "will watch the recording" (they won't); and the decisions announced in minute thirty-four exist officially nowhere — until someone asks in Slack three weeks later and four people paraphrase four versions.

Meeting notes solve this and nobody writes them for all-hands, because an hour-long town hall is miserable to minute manually. So the company's most important recurring communication produces its least reliable record: vibes, differentially remembered.

The mechanics

How veridive MCP helps

library_import_youtube_video archives each all-hands as it's posted — most companies already upload them unlisted; the import adds the transcript and makes the archive queryable. run_youtube_tool's meeting-notes and action-items do the miserable part: structured notes with the decisions, announcements and owners extracted, from what was actually said.

Twenty minutes after the recording posts, the canonical record exists: notes for the skimmers, action items for the owners, and a searchable archive for the three-weeks-later Slack question — answered with a link to the minute instead of four paraphrases. The same pipeline covers every recorded recurring meeting the company runs.

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
    Import the recording

    Tagged by meeting and date — the archive builds itself meeting by meeting.

    Import this all-hands recording into my library, tagged 'all-hands <date>': <video URL> — use veridive's library_import_youtube_video.
  3. 3
    Generate the record

    Meeting notes plus extracted action items — the canonical version, twenty minutes after upload.

    Generate meeting notes and action items from the imported all-hands — decisions, announcements, owners, with timestamps.
  4. 4
    Serve the archive

    The three-weeks-later questions get answered from the record, not from memory.

    Search my all-hands archive: when did we announce <decision/change>, and what exactly was said? Cite the meeting and minute.
Before vs after

Problems it solves

  • All-hands decisions live in differential memory and Slack paraphrase wars.
  • Nobody minutes an hour-long town hall manually, so the record is vibes.
  • 'Watch the recording' is a promise, not a behavior — non-attendees stay unsynced.
  • Finding when something was announced means scrubbing months of recordings.
The payoff

Why it's valuable

  • A canonical record twenty minutes after upload — notes, decisions, owners, timestamps.
  • Skimmers read five minutes; the archive answers the retroactive questions with links.
  • Action items stop evaporating between the announcement and the follow-up.
  • One pipeline covers every recorded recurring meeting, not just all-hands.
Copy & paste

Example prompts

Import this all-hands, tagged 'all-hands <date>': <URL> — use veridive's library_import_youtube_video.
Meeting notes + action items with owners and timestamps.
When did we announce <change>, and what was said exactly? Cite meeting and minute.
Questions

FAQ

Our all-hands recordings are unlisted — does that work?

Unlisted YouTube videos import by URL like any other. Private videos don't — for those, the recording needs to be accessible to the importing account's view.

Which tools produce the notes?

run_youtube_tool's meeting-notes for the structured record and action-items for the extraction — both running on the imported recording. The archive layer is the library itself.

Is this a Circleback/meeting-assistant replacement?

Different layer: meeting assistants live in the call; this processes the published recording and — crucially — builds the searchable archive across months of meetings, which call-time tools don't.

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.