Margin Call
Crypto

About

Last updated: August 26, 2026

AI-powered analysis automation tool

Margin Call is not human-written investment advice — it is an automation tool that combines a large language model (Google Gemini) with deterministic rule-based logic to organize and summarize public market data. All analysis is auto-generated information for reference and is not investment advice.

What is Margin Call?

Margin Call is a free web and mobile tool that takes US and Korean stock market quote data, automatically analyzes and summarizes it with AI models, and presents it through an interface. It automates everyday investment-research chores — stock search, automatic commentary on key financials, AI analyst reports, sector trend commentary, portfolio P&L tracking, exit-strategy management, and AI-powered Discovery scans.

Use it right in your browser with no installation required; on mobile, add it to your home screen as a PWA and it behaves like a native app. There are no usage or subscription fees — it is funded entirely by advertising.

What it really is — an "automation tool," not "investment advice"

Let's make this clear up front. Margin Call is not a publication run by human analysts, and none of its output evaluates an individual investor's financial situation, investment objectives, or risk tolerance. As such, this service does not constitute a financial investment or investment advisory business under capital markets law, and holds no related license.

The value this service provides is limited to three things: (1) information aggregation — gathering scattered public market data onto one screen; (2) information organization and summary — automatically explaining standard financial metrics; and (3) a personal tool that automatically calculates the P&L and sector allocation of the holdings and trades you enter yourself. None of the output is a buy/sell recommendation.

AI usage spec — where, which model, and how

For transparency, we disclose every place AI is used in the service and how.

Automatic stock commentary (/stock/[symbol])
Rule-based (no LLM)
Purpose: Automatically converts key financial metrics like P/E, P/B, ROE, and debt ratio into sentences by comparing them against predefined thresholds
Guardrail: Deterministic output. The same inputs always produce the same sentences. It contains no market outlook from the operator and no buy/sell recommendation
AI analyst report (modal)
Google Gemini
Purpose: Summarizes company overview, peer comparison, valuation context, and major risks based on public data
Guardrail: Output always shows an 'AI-generated' label and disclaimer. Prompt guardrails prevent it from answering without source data
Sector commentary (/sectors)
Google Gemini
Purpose: Summarizes the daily move patterns of 11 GICS sector ETFs in natural language, with commentary on fund flows and market sentiment
Guardrail: Only ETF price data is fed in. Individual stock recommendations and future-direction forecasts are blocked at the prompt level
Portfolio sector outlook (/sectors)
Google Gemini
Purpose: Takes your portfolio's sector weights plus market data and generates diversification commentary and general observations
Guardrail: No personally identifiable information is sent to the model. Calls are limited to once per day. The same weight pattern reuses a cached result
AI Discovery scan (/discovery)
Google Gemini + quantitative scoring
Purpose: Quantitatively computes four scores — momentum, valuation, financials, and catalyst — then generates a summary of the top candidates with AI
Guardrail: The scoring itself is a deterministic formula. AI only handles the final explanation. All output is a 'discovery candidate,' not a buy recommendation
Exit analysis (/exit)
Google Gemini
Purpose: Generates commentary to review sell timing based on the target price, stop-loss, and current holding you entered
Guardrail: Only called when you explicitly press the button. No automatic triggers. Labeled 'AI opinion'
Glossary (/learn/*)
No LLM
Purpose: 47 learning guides on investing terms, strategies, and metrics
Guardrail: Static content written by the operator with reference to standard financial-analysis textbooks. Not AI-generated

The track record is published, not curated

Every pick from AI discovery and early signals is written to an append-only ledger the moment it is generated, and is never edited or deleted afterwards. There is no survivorship bias from quietly removing losing picks. Each pick's 5- and 20-trading-day return is scored as alpha — the excess return after subtracting that market's benchmark index over the same period — so a pick that merely matched a rising index scores zero.

The full ledger is public with no login and no payment. You can see the average and median alpha by market and by path, the win rate, and the most recent picks as they stand — the median is published next to the average so you can tell when a few large winners are carrying it.

See the AI pick track record

Limitations of AI output (please be aware)

  • It can be factually wrong. Given the limits of training data and the nature of the models, an LLM can produce output with factual errors (commonly called "hallucination"). Trust the quantitative quote and financial figures pulled directly from the source API, and treat AI-generated natural-language explanations as reference only.
  • It may not reflect the latest information. Events after the model's training cutoff (earnings releases, M&A, regulatory changes, etc.) are not automatically reflected in AI output. Always cross-check official disclosures and news before any important decision.
  • Bias can exist. Depending on the distribution of the LLM's training data, US market material may be more richly represented than Korean market material, which can lead to differences in the depth of output.
  • It is not individual advice. AI output is generated without knowing your financial situation, investment objectives, or risk tolerance. You must make actual investment decisions yourself, based on your own circumstances.

Data sources

  • Quotes & financial metrics — Public Yahoo Finance API. May be delayed 15-20 minutes versus real-time exchange quotes; after the close it reflects the closing price.
  • Korean stock classification & company-name mapping — Curated in-house based on KRX (Korea Exchange) filings. We normalize market classification (KOSPI/KOSDAQ) and company names.
  • Sector classification — Uses the prices of the 11 GICS (Global Industry Classification Standard) sector ETFs (XLK, XLF, XLE, XLV, XLI, XLC, XLY, XLP, XLU, XLRE, XLB).
  • Exchange rate — The KRW/USD rate is the daily average from a public FX API.
  • Crypto quotes — Public CoinGecko API.
  • AI analysis — Google Gemini API. Input is limited to market data fetched from the public APIs above.

Personal data and AI calls

When calling AI models, we do not send your personally identifiable information (email, account identifiers, etc.). The input sent to the AI is limited to market/portfolio data such as stock symbols, quote and financial figures, and sector weights. User accounts and AI call logs are managed separately. For details, see the Privacy Policy.

Key features at a glance

  • Watchlist — Organize hundreds of stocks into groups and monitor price, change, and sparklines.
  • Portfolio management — Record holdings and trades, auto-calculate average cost, support mixed KRW/USD portfolios, and auto-compute sector allocation.
  • AI analyst report — Company overview, earnings, peer comparison, valuation, and major risks, auto-organized with Gemini.
  • Exit strategy — Set target and stop-loss prices, trailing stops, and scaling-out rule templates.
  • Sector dashboard — 11 GICS sectors + 21 curated themes, with your portfolio's leading sector vs. the market average.
  • AI Discovery — Automatically scans the entire market by momentum, valuation, financial, and catalyst scores.
  • 47 learning guides — Guides spanning valuation, financial analysis, technical analysis, trading strategy, macro, and portfolios.

Limitations of the data & tool

Quote data may be delayed about 15-20 minutes versus real-time exchange quotes, and after the close it reflects the closing price. This service does not provide brokerage or order-execution functionality; actual buying and selling must be done directly in your own brokerage system.

Revenue model (advertising disclosure)

Margin Call is a free service funded by advertising. There is no premium subscription or paid plan, and all features are available equally to all users. This page and some content pages may display Google AdSense and Google AdMob ads.

Ads may be shown based on your interests, and for this Google AdSense may use cookies and an advertising ID. You can disable ad personalization in Google Ad Settings. This service has no separate advertising contract or affiliate relationship with any specific stock, brokerage, or financial product, and content is written separately from ads.

Ad revenue is used for quote API fees, AI model call costs, and infrastructure upkeep.

Investment notice

All information provided by this service (quantitative quote data, AI auto-generated analysis, rule-based commentary, and all learning content) is reference material for investing and does not recommend buying or selling any specific stock or guarantee returns. Stock investing carries the risk of losing your principal, and you are solely responsible for all investment decisions and their outcomes.

Neither this service, its operator, nor any automated system it uses (including AI models) evaluates an individual investor's financial situation, investment objectives, or risk tolerance, and therefore does not provide individual investment advice. Before any important investment decision, be sure to also use official financial statements, company disclosures, and consultation with a licensed financial advisor.

Contact

For all inquiries about service operation, feature suggestions, bug reports, and privacy, contact us at kyd3534@gmail.com. You can also report factual errors in AI output (which sentence on which stock's page is wrong) to the same address and we'll review it promptly.

Quick links