Options trading MCP server
Connect Claude or ChatGPT to SPX 0DTE options backtests.
Cashflow Engine’s MCP connector is a remote Model Context Protocol server for options traders. It lets Claude, ChatGPT, Cursor and other supported MCP apps query a database of 275,000+ backtested SPX 0DTE strategies, build portfolios and run Monte Carlo simulations in plain English. Every result is a hypothetical backtest.
https://mcp.cashflowengine.io/mcpIncluded on every plan, Blueprint included. Nothing it does touches a broker.
Watch an AI run a real 0DTE backtest.
A replay of a real conversation: the AI calls the connector, runs a Hunter search and a Monte Carlo — and flags the result that looks too good to be true.
Real conversation · Claude + Cashflow Engine connector
The dashboard builds here once the conversation reaches the last stage.
Runs the real engine
Claude, ChatGPT, Gemini or Grok runs the Hunter, portfolio analytics and Monte Carlo through the connector.
Tells you what’s wrong
It pushed back on its own first result.
Builds what you ask for
Dashboards, charts, comparisons.
Works with Claude, ChatGPT, Gemini, Grok and any MCP-compatible assistant. Available on all plans.
Backtested / simulated conversation and dashboard. Hypothetical backtest results. Not a forecast. Not financial advice.
What is MCP?
MCP (Model Context Protocol) is an open standard that lets an AI assistant call outside tools. Instead of only answering from what it was trained on, the AI can ask a connected service for real data and use the answer. For options trading, that means your AI can query Cashflow Engine's backtests directly instead of guessing at numbers.
You don't need to know anything about AI, code or prompts. If you can describe a strategy to another trader, you can describe it to your AI.
- 1
You ask
In normal words, with your own constraints: strategy type, period, drawdown limit, how many strategies.
- 2
The AI picks a tool
It translates your question into a call to one of the connector’s tools — strategy search, portfolio analysis, Monte Carlo.
- 3
The engine runs it
Cashflow Engine answers from its backtest database — the same numbers you’d see in the Workbench.
- 4
You get the answer
The AI summarizes the result, draws a chart or table if you ask, and you follow up from there.
How to connect Claude or ChatGPT in 3 steps
- 1
Add the connector
In your AI app, add a custom connector (sometimes called an MCP server or remote tool) and paste the server URL above.
- 2
Sign in and approve
You'll be asked to sign in with your Cashflow Engine login and approve the connection. It only ever acts as you, in your own account.
- 3
Ask in plain language
Ask your AI to search strategies, build portfolios, run a Monte Carlo, or explain a result — it calls the same engine the Workbench uses.
Some apps also ask for a client ID — the exact per-app steps are in the Workbench under Connect your AI.
Works with supported MCP-compatible AI apps, including
- Claude (web/desktop)
- Claude Code
- ChatGPT
- Codex
- Cursor
- Gemini CLI
What you can ask: 0DTE strategy backtesting in plain English
You set every constraint — the strategy family, the period, the thresholds. The AI runs the engine and shows you what the data says.
Strategy search: MEIC, METF, iron condors and credit spreads
Filter the SPX 0DTE catalog the way you would in the Strategy Browser — by structure, entry time, premium, width and stop.
“Show MEIC strategies with MAR above 2 over the last 26 weeks, no Fridays.”
“List SPX same-day-expiry iron condors with a 50-point width, excluding FOMC and CPI days, sorted by max drawdown.”
“Compare METF put credit spreads entered at 10:33 vs 14:00 — win rate, P/L per contract and max drawdown.”
Portfolios, correlation and drawdown
Combine strategies you chose and measure them as one portfolio — or let the Hunter search under rules you write.
“Show the correlation matrix of these four strategies.”
“Which days drove the drawdown in this portfolio?”
“Run the Hunter for 6–10 strategies with a combined max drawdown under 10% over the last 26 weeks.”
Monte Carlo stress tests
Resample a portfolio’s daily P/L into thousands of paths and look at the spread of outcomes, not just the one backtest line.
“Run a Monte Carlo on my saved portfolio X.”
“What were the median and 95th-percentile drawdowns across 5,000 Monte Carlo paths for this portfolio?”
Explaining the results
No finance degree or AI background needed — ask what a number means and the AI explains it in plain English.
“Explain what MAR and Confluence mean for this strategy, in plain English.”
“Why does this strategy have a high win rate but a small P/L per trade?”
AI options backtesting: why the numbers come from the engine
The AI does not run backtests or invent figures — it reads them from Cashflow Engine. Every strategy in the catalog has already been backtested on second-level SPX options data, and the whole grid of 275,000+ variants is re-scored every week. When you ask a question, the connector returns those stored results, so an answer takes seconds and matches what you would see in the Workbench.
What the AI adds is the conversation: turning your question into the right filters, putting several results side by side, and explaining them. It can still misread a result, so anything you act on is worth checking in the Workbench itself. The full data and fill methodology is in the Data & methodology docs.
What the connector can and can't do
It can
- Search and filter the backtested SPX 0DTE strategy catalog.
- Analyze a portfolio: equity curve, drawdown, correlation, trade log.
- Run the Hunter under the rules you set, and Monte Carlo simulations.
- Save a portfolio to your own account — only when you ask it to.
It can't
- Connect to your broker, or place an order of any kind.
- Move money, withdraw, transfer, or touch funds in any account.
- Turn a strategy on. A saved portfolio stays inactive until you open it yourself in the Workbench.
- Tell you what to trade. It describes what the data shows — you set every constraint and make every decision.
It runs under your own login and only sees your own saved portfolios and data. Every number it shows you is a hypothetical backtest or simulation, not a live result or a forecast.
MCP connector FAQ
What is an options trading MCP server?
An MCP (Model Context Protocol) server is a connector that lets an AI assistant call an outside tool. Cashflow Engine’s MCP server gives Claude, ChatGPT and other MCP-compatible apps access to its SPX 0DTE backtest database, portfolio analytics and Monte Carlo simulation, so the AI answers from real backtest data instead of guessing.
Do I need to know anything about AI to use it?
No. You type questions the way you would ask another trader, for example “show MEIC strategies with MAR above 2 over the last 26 weeks”. The AI chooses the right tool and filters; there is no code, no prompt syntax and nothing to install.
What data does it use?
The same Cashflow Engine backtest database as the Workbench: 275,000+ SPX 0DTE strategy variants — MEIC, METF, ratio spreads and more — backtested on second-level options data, with the whole grid re-scored every week.
Are the results real trades?
No. Every number it returns is a hypothetical backtest or a Monte Carlo simulation, never a live fill or a forecast. Past performance does not predict future results.
Can I describe a strategy in my own words?
Yes. The AI maps phrases like “entries after noon”, “no Fridays” or “skip FOMC days” onto the catalog’s filters — entry time, width, premium, stop, weekday and event days. When a phrase could mean more than one setting, it is designed to ask you rather than guess.
Which plans include this?
Every active Cashflow Engine plan — Blueprint included. There is no separate add-on for it.
Does it cost extra?
No. The MCP connector is included with your subscription at no additional charge.
Is my data shared with other users?
No. It runs under your own Cashflow Engine login and only ever sees your own saved portfolios, backtests, and account data — never another user’s.
How do I disconnect it?
Remove the connector in your AI app’s own settings (most apps list it under connectors or integrations). You can also revoke access any time from the Workbench, under Account → Connected apps.
Which AI app do I need?
An app that supports remote MCP connectors. Supported today: Claude (web and desktop), Claude Code, VS Code, ChatGPT, Codex, Cursor and Gemini CLI. Claude, Claude Code and VS Code only need the server URL; the others also need a client ID shown in the Workbench under Connect your AI.
Can it place trades or move money?
No. It can read, analyze, backtest and simulate — the same things you can do by hand in the Workbench. It cannot connect to a broker, place an order, move funds, or turn a strategy on. That always happens when you open the Workbench yourself.
No account yet?
The connector comes with every plan. Start the 14-day free trial, then paste the server URL into your AI app.
Hypothetical backtested performance. Past performance does not predict future results. Cashflow Engine is analytics software, not an investment adviser, and nothing on this page is a recommendation to trade.