MCP

AI-agent format

Turn central-bank speeches into a macro brief

Batch-summarize policy speeches and pressers into one macro memo you can actually finish before the open.

CFO / AnalystFinance, Investing & Due Diligencedeep_searchrun_youtube_tool
The use case

The problem

Central bankers publish their thinking out loud, on camera, constantly: the presser after the decision, the testimony before the committee, the speech at the symposium, the panel where a regional president ad-libs past the prepared text. For anyone whose costs, hedges or portfolio ride on rates, it's the highest-stakes primary source there is — and it arrives as hours of deliberately measured speech per week.

The market's solution is to outsource the listening: read what a journalist heard, filtered through a headline desk. Which works, on average, until the week it doesn't — when the shift was in the Q&A phrasing everyone summarized past, and the people who watched the tape repriced before the people who read about it.

The mechanics

How veridive MCP helps

deep_search collects the week's tape: speeches, pressers and testimony, recency-filtered so you're always working the current docket, with the news/lecture content types keeping commentary channels out of the primary-source pile. Evidence timestamps mark where the topics you track — inflation language, labor read, balance-sheet talk — come up in each appearance.

run_youtube_tool does the compression: summarizer per appearance for the structured brief, key-quotes for the exact phrasings — because in this game the unit of signal is the sentence, and "somewhat elevated" versus "elevated" is a trade. The output assembles into a morning-ready macro memo where every observation cites its moment on tape, built from the primary source instead of the paraphrase market.

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
    Collect the week's docket

    Recency-filtered sweep of policymaker appearances — the decision presser plus the speech circuit around it.

    Find this week's <central bank> speeches, press conferences and testimony — full-length primary sources, newest first — use veridive's deep_search.
  3. 3
    Compress to briefs and phrasings

    Summaries for structure, key quotes for the exact language — the sentences the market will parse.

    Summarize each of these appearances and extract the key quotes on inflation, the labor market and the policy path.
  4. 4
    Assemble the memo

    One page: what was said, what changed versus the last round, which phrasings moved. Every line cited.

    Draft a macro brief from these summaries: main messages, language changes versus recent appearances, notable Q&A moments — with citations.
Before vs after

Problems it solves

  • Hours of deliberately slow, hedged speech per week — unwatchable live, dangerous to skip.
  • Headline summaries flatten exactly the phrasing gradients ('somewhat', 'likely', 'in coming meetings') that carry the signal.
  • The unscripted Q&A — where the real information density lives — is the part coverage summarizes worst.
  • Cross-speaker synthesis (governor vs regional presidents diverging) requires watching everything, so nobody does it.
The payoff

Why it's valuable

  • A primary-source macro brief before the open, instead of secondhand takes after it.
  • Key quotes preserve exact phrasing — the unit of signal in policy communication.
  • Language deltas versus prior appearances surface the drift that headlines call 'no news'.
  • The same pipeline covers any policymaker circuit: domestic, ECB, BoJ — one workflow, every docket.
Copy & paste

Example prompts

This week's <central bank> speeches, pressers and testimony, newest first — use veridive's deep_search.
Summarize each and pull key quotes on inflation and the policy path.
Macro brief: main messages, language changes vs recent appearances, notable Q&A — cited.
Questions

FAQ

Why not just read the published speech texts?

The prepared text is the least informative layer — it's the Q&A and pressers, unscripted and unpublished-as-text, where phrasing shifts first. Video transcripts cover exactly the layer the PDFs miss.

How fast after an appearance can this run?

As soon as the video is up and transcribable — for major pressers, typically well within the same session. The recency filter (today / this week) keeps the sweep on the live docket.

Is this investment advice?

No — it's a briefing pipeline over public primary sources. What the phrasing means for your rates exposure is your call; this just guarantees you heard it first-hand.

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.