Quant Investing MCP commands

Quant Investing MCP commands: the plain-English command library for Claude

A short, tested list of phrases that get good answers from the Quant Investing connector in Claude. Copy one, change the details, and go.

By Tim du Toit, founder of Quant Investing. 39 years investing since 1987, author and published quant researcher. Last updated: June 2026.

Quant Investing MCP commands, definition:
The plain-English questions a Best Deal subscriber types into Claude AI to drive the Quant Investing Model Context Protocol (MCP) connector. One connection gives Claude access to the live stock screener (24,000 companies, 200+ factors), the Quant Value newsletter archive (since 2010), the Shareholder Yield Letter archive (since May 2023), the subscriber's own watchlists and saved screens, and a full backtesting engine, through 32 callable tools. You do not need special syntax. You ask in normal English, Claude picks the right tool.

You do not need to memorise anything. The whole point of the connector is that you ask the way you would ask a sharp analyst sitting next to you. But a few phrasings work better than others, and I have collected the ones I reach for most below.

Two habits make almost every answer better.

Be specific about the universe. "European stocks", "US small caps", "market cap above 1 billion". Narrowing the field gets you a focused list instead of the whole world.

Say what to sort by. "ranked by earnings yield", "highest dividend first", "best Piotroski score". If you skip this you often get the largest companies first, which is rarely what you wanted.

Key facts about the Quant Investing MCP connector:
• 32 callable tools across the screener, both newsletters, your own data, and the backtest engine.
• Database: 24,000 listed companies, 200+ financial factors, refreshed daily.
• Daily limit: 500 tool calls per key. Backtests count as 5. Repeat cached answers are free.
• Access: Best Deal subscription only (Screener, Quant Value newsletter, Shareholder Yield Letter). 79.70 EUR per month or 779.70 EUR per year.
• Setup time: about 5 minutes. See the step-by-step setup guide.

1. Find stocks (the screener)

This is what most people connect for. Ask for a filtered, sorted list and Claude runs the screen across the full universe.

• "Show me the top 50 stocks by ROIC in Europe with a market cap above 500 million."
• "Find UK stocks with an earnings yield above 10% and a Piotroski F-score of at least 7."
• "Cheap, high-quality stocks: QI Value score in the best quartile and F-score 8 or 9, in North America."
• "Show me Japanese stocks trading below book value with positive momentum."
• "Which energy companies have a shareholder yield above 8%?"

Not sure what a factor is called? Just ask.

• "What screening factors are available for dividends?"
• "List the stock universes and how many companies are in each."

2. Look at one company

When you want the full picture on a single name rather than a list.

• "Give me the full profile for Apple."
• "What is Bayerische Motoren Werke trading at, and how do its earnings yield and ROIC compare to its sector?"
• "Show me everything you have on ISIN DE0005407407."
• "Search for companies with 'Sainsbury' in the name."

3. Prices and currency

• "Show me the daily prices for Shell over the last two years."
• "What is 50,000 USD worth in EUR today?"
• "Convert 1,000 GBP to EUR as of 1 March 2024."

4. Predefined strategy screens

These are the ready-made strategies from the website screener: Magic Formula, ERP5, O'Shaughnessy's Tiny Titans, Dividend Quality, Negative Enterprise Value, Quality Small Companies, and around 20 more. You can run any of them straight away.

• "Show me my saved screens." This also lists every predefined strategy with a short description.
• "Run the Magic Formula screen and show me the top 20."
• "Run O'Shaughnessy's Tiny Titans screen, top 10."
• "Run the Dividend Quality screen and show me the top 15."
• "Run the Negative Enterprise Value screen."

Ask for a name that does not exist and Claude shows you the full list of the ones that do, so you are never left guessing.

One thing to know about results. A predefined screen returns exactly what the website screener returns, including the quirks of how each screen is defined. A screen sorted by trailing dividend yield, for example, can show a company paying a one-off special dividend right at the top. That is the screen definition at work, not an error.

5. Your own saved screens

• "Show me my saved screens."
• "Run my 'Quality Cheap' screen and show me the top 30."
• "Which of my saved screens don't have a sort order set?" Useful before you backtest them.

6. Your watchlists, alerts and comments

• "List my watchlists."
• "Show me the stocks in my 'Dividend ideas' watchlist with their current prices and ratios."
• "What price alerts do I have, and which have triggered?"
• "Show me my notes on individual stocks."
• "What grid or column templates have I saved?"

7. The newsletters: current portfolio and picks

Works for both the Quant Value newsletter (small cap) and the Shareholder Yield Letter (large cap).

• "Show me the current Quant Value portfolio with live prices."
• "What are the active Shareholder Yield picks in Europe?"
• "What did Tim recommend in the June 2026 Quant Value issue?"
• "What did Tim sell this month in Shareholder Yield?"
• "Give me the full analysis for the CeoTronics pick."
• "Search the newsletters for anything about palm oil."
• "What did the latest editorial say about the worst drawdown in the newsletter's history?"

8. The newsletters: performance

• "How has the Quant Value portfolio performed since inception, in EUR?"
• "Show me the return of each Shareholder Yield pick: entry price, current price, dividends, total return."
• "Which Shareholder Yield picks overlap with my 'Dividend ideas' watchlist?"

9. Backtesting

This is the most useful part of the connector. Three different questions, three tools, and Claude picks the right one. If you want to understand the method behind it, read how to back test your investment strategy.

A whole newsletter, as you would actually have traded it:

• "Backtest the Quant Value newsletter from July 2010 in EUR with 100,000 starting capital, ending 31 December 2025. Show me the return, CAGR, volatility and worst drawdown."
• "Backtest the Shareholder Yield portfolio since its first issue in May 2023, in EUR."
• "What if I had sized every position at 5% instead of 2%?"

One of your saved screens, rolled forward over many years:

• "Backtest my 'Quality Cheap' screen on a 12-month rolling basis from 2018 to today, sorted by earnings yield, and compare it to the universe."
• "Run that again but tell me the worst peak-to-trough drop and when it happened."

A single snapshot, what a screen would have picked on one date:

• "What would my screen have picked on 2 January 2020, and how did those stocks do over the next 12 months?"

A worked example, so you know what the output looks like. A backtest of the Quant Value newsletter from 15 January 2020 to 6 May 2026, starting with 100,000 EUR and dividends reinvested, returned a total of 149.02% (a compound annual growth rate of 15.57%), with a maximum drawdown of 6.66%. Over that run the portfolio held 47.6% of ideas at a profit and 52.4% at a loss. The point of the strategy is that the winners are larger than the losers, not that every position works. You can reproduce the same number in your own account.

A few things worth knowing about the engine:

• The benchmark is the average of the screen's own investable universe, not the MSCI World or the S&P 500. Those indices are not in the data.
• Returns include reinvested dividends.
• Start a backtest before about 2015 and Claude reaches into the deeper price archive for the early years, then tells you how many picks it covered and how many it had to skip. No silent gaps.

10. Account, quota and speed

• "What's my quota status?" Free to ask, does not use up your daily allowance.
• "How warm is the cache?" Also free. Shows what has been cached for you.

Heavy backtests count as 5 of your daily calls each, so a 500-a-day allowance is roughly 100 backtests. Ask the exact same backtest twice and the second one returns almost instantly from the cache, free for 7 days.

Here is how the daily quota works.

11. If something looks off

• The screener filter returns nothing. Check the country name and the ranges, then ask Claude to relax one filter at a time.
• A backtest return looks far higher than expected. Ask Claude how many picks had a one-off special dividend. Those can inflate a single period.
• A backtest return looks far lower than the screen's intent. Make sure you told Claude what to sort by. Without it the screen ranks by size, which usually is not the point.
• A factor is not found. Ask "is this factor available for live screening, for backtesting, or neither?" A few factors work for today's data but not for history, and Claude will tell you which.

How the MCP connector compares

MCP connector versus a traditional API. A normal API returns raw data for a developer to parse with code. The MCP connector exposes the same data as 32 structured tools that Claude can call during a chat. You write no code. You ask in English and read the answer.

Quant Investing versus other screening services. No other service currently expose an MCP connection to Claude or any other AI model, and none of them put a 16-year newsletter archive next to live screener data in one query. The combination of live data, the editorial archive, and a working backtest engine in a single conversation is what makes this connector different.

This page is general information about a research tool, not personalised investment advice. The example commands are illustrations of how the connector works, not recommendations to buy or sell any security. Prices move, and you can lose money investing. Past performance is no guarantee of future results. See our full disclaimer.

Claude Screener MCP Frequently asked questions

What are the Quant Investing MCP commands?

They are plain-English questions you type into Claude AI once the Quant Investing connector is added. There is no command syntax to learn. You ask in normal language, for example "Show me the top 20 European stocks by F-score above 7", and Claude calls the right one of the 32 tools and formats the answer.

Do I need to learn special commands or code?

No. The connector is built so you ask the way you would ask a person. The phrasings on this page are starting points you can copy and change, not a fixed language. Writing the universe (where) and the sort order (ranked by what) into your question is the only habit that reliably improves answers.

Which subscription do I need to use these commands?

The Best Deal subscription, which bundles the Quant Investing screener, the Quant Value newsletter, and the Shareholder Yield Letter. It costs 79.70 EUR per month or 779.70 EUR per year. The connector is not sold separately. A screener-only or newsletter-only subscription does not include the API key you need to connect.

How many commands can I run per day?

500 tool calls per key per day, reset at 00:00 UTC. Standard tools cost 1 unit. The four backtest tools cost 5 units each, so the allowance is roughly 100 backtests a day. Quota and cache checks are free, and a repeat backtest that hits the cache is free for 7 days.

What can I ask the backtest engine?

Three kinds of question. Backtest a whole newsletter as you would have traded it, roll one of your saved screens forward over many years, or take a single snapshot of what a screen would have picked on one date. Returns include reinvested dividends, and the benchmark is an equal-weighted basket of the screen's own investable universe rather than a market index.

Can I run my own saved screens?

Yes. Ask "Show me my saved screens", then "Run my X screen and give me the top 30". You can also ask "Which of my saved screens don't have a sort order set?" before you backtest, because a screen with no sort order quietly ranks by size.

How is this different from a normal stock screener API?

A normal API hands raw data to a developer who writes code to use it. The MCP connector hands 32 structured tools to Claude, which calls them for you inside a chat. You never touch code. You also get the screener, both newsletter archives, your own watchlists, and the backtest engine through one connection rather than separate endpoints.

Is my data private when I use the connector?

Yes. Your API key is bound to your subscriber account and stored only as a hashed value on the server. Your personal data (watchlists, alerts, comments, saved screens) is scoped to your account, so no other subscriber can see it through the connector.

How do I set up the connection?

It takes about 5 minutes. Generate an API key on your subscription page, add Quant Investing as a custom connector in Claude.ai, and ask your first question. The full walkthrough with screenshots is in the step-by-step setup guide.