What I Built

StockPulse is an AI-powered Indian Stock Intelligence platform built entirely on Notion.

It solves the problem of scattered financial data by acting as a centralized, human-in-the-loop research hub. First, a Python data pipeline fetches daily price and delivery data from the NSE and BSE (Indian stock exchanges). It then runs 5,000+ stocks through a rigorous, battle-tested 12-condition fundamental screener.

The magic happens when the data enters Notion. By using the Model Context Protocol (MCP), StockPulse allows AI assistants (like Claude) to seamlessly read the screened data, identify anomalies, analyze fundamentals, and write comprehensive research reports directly back into Notion databases.

Show us the code

Safvan-tsy / stockpulse

Using Notion MCP to Screen 5,000+ Stocks and Write AI Research Reports

StockPulse India 📈

AI-Powered Indian Stock Intelligence on Notion — Built for the Notion MCP Challenge.

StockPulse takes daily price and delivery data from NSE & BSE, screens 5000+ stocks through 12 battle-tested fundamental conditions, and uses a dual-MCP architecture — the official Notion MCP for workspace I/O plus a custom StockPulse MCP for domain computation — to generate research reports, detect anomalies, and maintain a smart watchlist — all centralized in Notion.

What It Does

Data Pipeline — Downloads BhavCopy + delivery data from NSE/BSE, or reads from pre-built Excel workbooks
12-Condition Screener — Filters stocks for: profitability (PE, EPS), growth (sales, profit YoY), governance (promoter pledging), financial health (debt/equity, current ratio, ROCE)
Notion as Single Source of Truth — 5 linked databases: Stocks Master, Daily Prices, Screener Results, Watchlist, AI Reports
Dual-MCP AI Intelligence — The official Notion MCP (https://mcp.notion.com/mcp) handles all Notion reads/writes…

View on GitHub

How I Used Notion MCP

Notion MCP is the core I/O layer of this project. The AI agent uses it for every interaction with the Notion workspace:

Reading data via Notion MCP:

notion-search and notion-fetch to find and retrieve stock pages from the Stocks Master database

query-a-database-view to fetch screened stocks, price history, and watchlist entries with filters

Writing results via Notion MCP:

create-a-page to publish AI-generated research reports into the AI Reports database

create-a-page to add stocks to the Watchlist database with notes and status

update-a-page to set AI Ratings (Strong Buy / Buy / Hold / Avoid) on stock pages

The custom StockPulse MCP server complements Notion MCP as a stateless computation engine — it contains zero Notion SDK calls. It receives stock data (fetched by the AI via Notion MCP) as JSON input and returns analysis results:

screen_stock: Applies the 12 screening conditions and computes a weighted quality score (0–100)

detect_anomalies: Identifies stocks with Piotroski ≥7, promoter holding changes, and high delivery %

compare_sector: Ranks a stock against sector peers on ROCE, PE, debt, and other metrics

generate_report_content: Formats analysis into structured markdown for the AI to save via Notion MCP

The workflow loop:

AI fetches data from Notion databases → Notion MCP

AI passes data to screening/analysis → StockPulse MCP

AI writes reports, ratings, watchlist entries back → Notion MCP

Human reviews in Notion UI, adds notes → AI reads them next cycle → Notion MCP

This creates a true human-in-the-loop system where the Python pipeline crunches the hard numbers, Notion MCP provides seamless workspace access, StockPulse MCP adds domain intelligence, and Notion organizes it all beautifully for the investor to review and act on.