MetaTrader 5's AI Assistant can now read Strategy Tester results

MetaQuotes has given the AI Assistant built into MetaTrader 5 direct access to the Strategy Tester. Build 6180, released on 3 September 2026, lets the assistant retrieve testing and optimisation results, check tester settings and read the tester log, and build 6230 on 24 September 2026 added tools to prepare an Expert Advisor's inputs for a test run.

What happened

The main AI features arrived over two months, documented in MetaQuotes' official release notes:

• Build 6060 (23 July 2026) added native Model Context Protocol (MCP) support and an AI Assistant in both the terminal and MetaEditor. External agents such as OpenAI Codex and Claude Code can connect to the terminal over MCP.
• Build 6180 (3 September 2026) extended the assistant's toolset to the Strategy Tester: it can retrieve testing and optimisation results, help configure optimisation criteria, check the current tester settings and read the tester, terminal and Expert Advisor logs.
• Build 6230 (24 September 2026) let the assistant read an Expert Advisor's input parameters and set them up for a tester run, launch and remove programs on charts, and query the Economic Calendar.

Build 6230 is the latest desktop release on MetaQuotes' release notes page at the time of writing.

The details that matter

You turn the AI features on by signing in with an MQL5.community account, which gives access to a free plan with MQL5.community as the default provider. According to the build 6060 notes, you can instead add your own API key for OpenAI, Anthropic, Gemini, DeepSeek, Ollama or another supported provider and pick the model in the terminal settings.

The same notes describe security settings that let you allow or block AI-initiated trading operations, or require manual confirmation, with separate controls for network requests and command line access. If you test EAs on the same terminal you trade from, set these before you let an agent run tasks on its own.

MetaQuotes also states that AI-generated responses are for information only and are not investment advice.

What it means for EA developers

The practical gain is speed. Instead of saving a report, opening it in Excel and scanning a long list of deals, you can ask the assistant to compare optimisation passes, explain why a run failed from the tester log, or confirm which symbol, period and inputs were actually used before you start a test. That last check alone catches a common mistake.

What hasn't changed is what the assistant is reading. It works from the tester's report and the tester's model of the market, and that model is only as good as its settings. MetaQuotes' tick generation documentation explains that generated ticks use a spread fixed for each bar, that "Open prices only" mode is fast but rough, and that stops and pending orders can fill at prices other than the ones specified in that mode. An assistant summarising a backtest built on those assumptions will summarise the assumptions too.

There is also a gap between a well-explained report and an edge. An assistant that helps set up and compare optimisation runs makes it easier to run many passes, which makes it easier to find a set of inputs that fits the past and nothing else. Faster reading of results doesn't replace out-of-sample tests, a realistic sample size or a forward test.

And the tester can't tell you how the EA behaves on your broker's real fills. The questions that decide whether an EA survives live are about its own trades in particular conditions: which sessions lose money, whether results fall apart when volatility rises, whether one weekday drags the rest down. Those answers come from live account history, broken down by session and condition, not from a summary of a simulated run.

What to do now

• Update the terminal and try the assistant on a tester report you already understand, so you can judge whether its reading matches yours.
• Ask it to confirm tester settings and modelling mode before each run, and prefer real ticks where your broker provides them.
• Treat optimisation it helps set up like any other optimisation: hold back data for out-of-sample checks.
• Export both the tester report and your live account history, and compare the two by session, weekday and volatility once the EA has enough live trades.

Read more

Our guide to exporting the MT5 Strategy Tester report covers saving the report, deals versus positions and exporting live history. The backtest validation checklist lists the checks to run before going live, and why a strategy works in backtest but not live covers where the two usually diverge. To see where your EA loses money by session, weekday, time and volatility, upload your MetaTrader 5 history (or trades from any supported broker, platform export or spreadsheet) for a free Entry Analysis report.