Intermediate · AI apps
Connect your AI assistant
Use Quantis Trade from Claude, ChatGPT or another AI app. It can search the recorded tests, explain a figure and estimate a price for free, and run a test on your tokens only if you allow it, with a cap on every run.
The problem
You may already ask an AI app about markets. It cannot run a backtest itself, and it can state a figure it has not looked up.
Connected to Quantis Trade, it calls tools that read stored results and run real tests through the same Strategy Agent as the website, and it is told the price before it spends anything.
Step by step
Add the address
In your AI app's connectors, add https://quantistrade.co.uk/mcp.
Sign in and choose
A Quantis Trade page shows what the app asks for. Reading is always included. Running tests, which spends your tokens, is off unless you tick it.
Ask in your own words
Ask it to search the recorded tests, explain a figure, or test an idea.
It checks the price first
Estimate a backtest's cost: Reads a plain-English strategy exactly as the Strategy Agent would and returns what would run (symbols, period, the rules as the result card will state them), the price in tokens and your balance, or the clarification or refusal the agent would give.
It runs within your cap
Every run carries the most it may spend. A run above that, or above your balance, is refused before anything is charged.
You
Test this with Quantis Trade: "Golden cross strategy on SPY over the last 10 years"
Your AI app
calls estimate_backtest_cost (free)Estimate a backtest's cost
calls run_backtest (within your cap)Run a backtest (spends tokens)
Replies with the verdict, the key figures and a link to the full result, and says what it spent.
A worked example from a stored test
An assistant asked to test a golden cross on SPY would send the Strategy Agent words like "Golden cross strategy on SPY over the last 10 years". A stored test from the truth matrix, the set of cases run against the engine on real prices to check its figures (case F01).
- You asked:
- Holds SPY while its 50-day average price is above its 200-day average, and stays in cash otherwise.
- Lost to simply holding SPY (S&P 500 ETF).
- Worst fall about the same: 34% against 34%.
- This rule
- $27,277
- from $10,000
- Simply holding
- $40,513
- from $10,000
- $10,000 put in at the close of 5 Oct 2016, counted at the close of 1 Oct 2026. An illustration: the test ran on 100,000 with fractional shares and percentage costs, so it scales to any amount.
- Only 4 round-trip trades in 10 years, so one decision can swing this result.
- A backtest of 6 Oct 2016 to 1 Oct 2026. It describes the past. Not a forecast, not advice.
- What each line means
- Full report
| Verdict | Lost to simply holding SPY (S&P 500 ETF). |
|---|---|
| Tested | 5 Oct 2016 to 1 Oct 2026 |
| This rule | $27,277 |
| Simply holding SPY | $40,513 |
| Starting amount | $10,000 (an illustration: a test scales to any amount) |
| What a run costs | 300 tokens, the same price as the website |
The same words sent today test up to today, so the figures would differ. The assistant is told the price before the run and what it was charged after it, and the run is kept in your Strategy Agent history like a run from the website.
What it cannot tell you
- Your AI app's provider receives whatever the tools return, under its own privacy terms. The tools return public evidence, Education Centre pages and your own results.
- Nothing the app does can trade, place orders, touch your pretend-money portfolios, change your settings or read another person's data.
- Tests use daily prices and at most 20 years, as on the website, and each key and app has rate limits.
- The app writes its own words around what the tools return. The figures come from Quantis Trade; the app's explanation of them is its own.
Try it
Once connected, type one of these in your AI app.
Use Quantis Trade to estimate what this test would cost: "Golden cross strategy on SPY over the last 10 years"
Search the Quantis Trade evidence for golden cross tests on UK shares.
Use Quantis Trade to explain what max drawdown means.
Related
- Feature guidesUse Quantis Trade from your AI assistantQuantis Trade runs an MCP server, so Claude, ChatGPT and other MCP clients can search our evidence, explain the numbers and run backtests for you, with a spending cap on every run.
- AI apps and APIAI apps and APIThe tools an AI app can call, from the list the server itself serves, and the limits on them. There is no general web API today.
- Pricing and tokensPricing and tokensWhat a test, a research question and the other paid tools cost, what the plans include, and what costs nothing.
- SecuritySecurity and data handlingHow account data, market data and connections are handled.
Tests describe the past. They are not forecasts and not advice. Read the risk warning