Skip to content
All projects

AI integration · Custom software

AI-Assisted Stock Market Assistant

An AI-powered chat app I built for a two-person household that answers questions in Turkish about Borsa Istanbul and US stocks using up-to-date data.

My role
Design and development · personal project
Status
In use · private

What need did it start from?

I wanted an assistant for the two people at home who follow the stock market: one you can ask about stocks in Turkish, with answers grounded in current data. The goal was to bring price, balance-sheet and news information that lives across different sites into a single conversation.

My approach

I didn’t want the model to just generate text; I wanted it to decide which data a question needs. Using the Claude API’s tool use, the assistant picks what it needs from tools for prices, history, financials, symbol lookup, the watchlist and web search, gathering data over several steps when necessary and combining it into one answer. I drew the boundaries early: the app only reads data. It never places orders, never connects to a brokerage account and gives no investment advice; it presents data and interpretation and leaves the decision to the user.

The solution

Prices, 52-week ranges, valuation ratios, fundamentals and analyst estimates come from Yahoo Finance (BIST data is delayed by about 15 minutes); for news, web search runs at most five times per answer. It can find a ticker from a company name, compare several stocks, keep an annotated personal watchlist and pick up earlier conversations. Costs are kept in check with prompt caching, token-based tracking and a daily spending limit. It works on phone and desktop and can be added to the home screen; access is login-protected, passwords are hashed and failed login attempts are rate-limited.

Where it stands

The app is in use in a two-person household. It isn’t public; only registered users can sign in.

Technologies

Application

  • Node.js
  • Express
  • Claude API (tool use, web arama, prompt caching)
  • Yahoo Finance
  • PostgreSQL / SQLite
  • HTML, CSS, JavaScript
  • Node test runner