Fintech Morning Briefing

A CLI tool that fetches articles from fintech, stablecoin, and MENA RSS feeds, scores each one with the Kimi K2 AI model, and saves the best picks to a Notion database. Run it every morning, then open Notion to draft your LinkedIn post.

Setup

npm install
cp .env.example .env

Fill in your API keys in .env (see below).

Run

node briefing.js
# or
npm start

Run time is typically 3–5 minutes depending on how many new items need scoring.

API Keys

Variable	Where to get it
MOONSHOT_API_KEY	Moonshot API console
NOTION_TOKEN	Notion integrations — create an integration, then share your database with it
NOTION_DB_ID	The 32-character ID in your Notion database URL (with or without hyphens)

Notion Database Setup

Create a database with these properties (exact names and types):

Property Name	Type
Title	title
Source	rich_text
URL	url
Score	number
Pillar	select
Format	select
Relevance Reason	rich_text
Suggested Angle	rich_text
Draft Hook	rich_text
Verification Needed	rich_text
Status	select
Date Added	date

Select options:

• Pillar: infrastructure, mena, commentary, none
• Format: Deconstruction, Contrarian, News Reaction, MENA Spotlight, Lesson, List, Schema
• Status: New (and any others you want for your workflow)

Configuration

Variable	Default	Description
MIN_SCORE	7	Only save items scoring at or above this (1–10)
MAX_ITEMS	60	Max items to score per run (controls API cost)

Lower MIN_SCORE to capture more items; raise it to be more selective. Lower MAX_ITEMS to reduce API spend on busy days.

Adding RSS Sources

Edit the RSS_SOURCES array at the top of briefing.js:

{ name: 'Your Source', url: 'https://example.com/feed.xml' },

Each source fetches up to 15 recent items. Failed sources are logged and skipped.

Seen URLs (seen_urls.json)

The script tracks URLs it has already scored in seen_urls.json (gitignored). On subsequent runs, those items are skipped so you do not pay to re-score the same article.

To reset and re-score everything, delete the file:

rm seen_urls.json

The file is capped at 2000 URLs; older entries are trimmed automatically.