Claude
- Open Add custom connector ↗
- Name it BacktestLoop and paste the URL
- Click Add, then sign in to your account
https://api.backtestloop.com/mcp
Backtest trading strategies
Describe a trading idea in plain English and BacktestLoop turns it into a real backtest on historical market data. When the run finishes, your AI shows the results with an equity chart, a drawdown chart and the key metrics.
With BacktestLoop you can:
Backtest a new strategy on stocks, ETFs, futures, forex or crypto, with your own rules for entries, exits, stops, targets and position sizing
Change an existing backtest, for example add a filter or a stop-loss, and rerun it
Compare variations side by side, such as different moving-average lengths or stop-loss levels, and keep the one you prefer
Stress-test a strategy on other years to see whether its edge holds up
Ask questions about a market's history, such as volatility by hour of the day or seasonal patterns
Turn a strategy explained in a YouTube video into a backtest
Find your earlier backtests and read their code, settings and logs, or start from one of BacktestLoop's example strategies
BacktestLoop writes, runs and repairs the strategy code on its own servers. Every backtest is saved in your BacktestLoop account, where you can open it for the full results, the trade list and the editable Python code.
Who it is for: traders and investors who want to test ideas on historical data before they act on them.
Limitations: BacktestLoop runs historical simulations only. In your AI it does not connect to brokers, place orders or give personal investment advice, and past performance does not guarantee future results. Results depend on the historical data available for each market. You need a BacktestLoop account; runs use that account's credits, the same as on the BacktestLoop website.
Backtest a 20/50-day moving average crossover on QQQ over the last 3 years
Follows the skill
https://api.backtestloop.com/mcphttps://api.backtestloop.com/mcpapi.backtestloop.comThe AI reads these server instructions before using this connector, every time. From the ChatGPT store listing.
BacktestLoop builds, runs and repairs real trading-strategy backtests in the user's BacktestLoop account (stocks, futures, forex, crypto). Route every request: new strategy -> ; change an existing backtest -> ; compare variations -> , then for the one the user picks; robustness on other years -> ; run again unchanged -> ; questions about market data itself -> ; a YouTube strategy video -> , then ; see a backtest or its results -> , shown with charts and tables as REPORTING A RESULT describes; see the code -> , then show it in a code block; stop a run -> . Never write strategy code yourself: describe the complete trading rules in plain language and BacktestLoop writes, runs and repairs the code. Do everything through these tools and never operate the BacktestLoop website with a browser or computer use; if no tool covers a request, say so plainly and give the backtest's BacktestLoop link. BacktestLoop only simulates trading on historical data: it cannot place orders, connect to a broker or trade live, and it gives no personal investment advice, so use no tool for such requests and say so (offering a backtest of the idea).
Defaults, never block on them: last 6 months, 100000 USD cash, the most natural timeframe (else 1h). Ask only when the symbol is truly unknown; call when unsure whether a symbol is supported. Build the user's rules exactly: never add or silently drop filters, stops, targets or timeframes.
Most runs finish within a few minutes; experiments and stress tests can take longer. Run tools return at once with a card that shows live progress. Then call with the ids from the run tool's result (workflow_run_id; for a rerun backtest_id; for a stress test backtest_id and batch_id) and keep calling it until it reports the run finished. Each call waits for up to about half a minute, so this takes a few calls; do not end your turn or reply before the run finishes, unless the user asked you not to wait. Then report the result as REPORTING A RESULT describes, and never state or guess a result you have not read. Do not call at the end of a run: a card shown after your report hides it, and the progress card already shows the result. If tells you to stop waiting, tell the user the run is still going, that the card shows its progress, and that they can ask you about it later. Never set up monitors, scheduled tasks or notifications for runs: their reports go to a separate conversation, never to this chat. When the user wants a run stopped, call with the same ids. Do not start the same request again. When the user asks about a run, or clicks Explain results, read the outcome first (, or ). Tool results may also list runs that finished since your last reply: report each as REPORTING A RESULT describes. At most two attempts of the same request per turn.
REPORTING A RESULT. You can render charts in your replies; a backtest report without charts is incomplete. When a build, change or rerun finishes, or the user asks to see a backtest's results, call with include_equity_curve, then answer in this order:
One sentence: what was tested and the headline result.
RENDER a line chart of the strategy's equity from chart.rows. Never skip it or replace it with a table.
Backtest trading strategies. Describe a trading idea in plain English and BacktestLoop turns it into a real backtest on historical market data. When the run finishes, your AI shows the results with an equity chart, a drawdown chart and the key metrics.
Open its listing in the app's connector directory and connect. Sign in to your account in the browser window that opens. It uses OAuth, so the AI never sees your password. Step-by-step setup for each app is on this page.
The BacktestLoop MCP server URL is https://api.backtestloop.com/mcp. Paste it as a custom connector in Claude, ChatGPT, Gemini or Grok, or add it to Claude Code, Cursor or any MCP client. Setup steps.
Yes. BacktestLoop is published or run by VECTORSPELL LTD, so it is the official MCP server. See other official connectors.
It is listed in 1 store: ChatGPT. Apps where it is not listed can usually still add it as a custom connector.
It is ranked 15,262 of 15,719 in the ChatGPT directory. A rank is the store's own order, not a usage count. How we track ranks.
It has 34 tools: 21 read data, 7 change data and 6 can delete or overwrite data. See every tool and its inputs.
Yes. 7 tools can change data and 6 tools can delete or overwrite it. Most AI apps ask before they run these. The store marks these as destructive:
It comes with 1 skill. A skill tells the AI how to do one job well with these tools. Read each SKILL.md.
Here are examples from its store listing. Type one into your AI after you connect.
Yes. Some of its tools show an interactive widget in the chat, not only text. See more connectors with interactive UI.
This page updates every day. To ask about BacktestLoop from Claude, ChatGPT or a coding agent, connect the Agent Ready MCP server: it tracks ranks, tools and new listings across the AI plugin stores.