Overview
What you can actually do with MTContext, in plain terms.
MTContext turns your MetaTrader 4 or 5 terminal into something an AI client can see and, if you allow it, act on. You install one Expert Advisor file on your terminal, connect an MCP client (Claude Desktop, Claude Code, Cursor, or anything else that speaks MCP), and from then on you can talk to your terminal in natural language.
What you can ask for
Prices and market data. "What's the current spread on XAUUSD?" "Show me the last 200 H1 bars for EURUSD." "What symbols do I have in Market Watch?" These map to read-only tools — no risk, no confirmation needed.
Account and position state. "What's my current equity and margin level?" "List my open positions." "What's my trade history for the last 30 days?"
Technical analysis. "Calculate RSI(14) on GBPUSD H4." "Where are the key support and resistance levels on US30 right now?" "Detect any chart patterns forming on the daily EURJPY chart." "What's the current relative strength of each major currency?"
Risk and sizing. "What lot size keeps me at 1% risk with a 30-pip stop on EURUSD?" "What's my correlation exposure across my open positions right now?" "Give me a risk summary."
Macro and sentiment context. "What's the interest rate differential between USD and JPY right now?" "Any high-impact news on the economic calendar today?" "What's the COT positioning on EUR futures?" "What's retail sentiment on GBPUSD?"
Trading — with a safety net. "Place a market buy on EURUSD, 0.1 lots, SL 50 pips, TP 100 pips." "Close position 12345." "Move my stop loss to breakeven on this trade." Every trade-affecting tool defaults to a dry run: it validates the request and tells you exactly what would happen, without touching the account. Nothing executes for real until you (or your AI client, on your explicit instruction) pass dry_run: false.
Equity protection. "Set an equity guard that closes everything if my account drops below $8,000." These run server-side, independent of whether your AI client is connected.
Coaching and journaling. "Explain why this trade would have worked or not, given the setup." "Backtest this strategy idea against recent price action." "Log a note on this trade for my journal." "What patterns show up in my trading behavior over the last month?"
The full list — every tool, its parameters, and what it returns — is in the Tool Reference, generated straight from the server's source so it's never out of date.
Prefer a REST API?
Everything above is phrased as MCP tool calls because that's how most people use MTContext — through an AI client. If you're building your own backend, script, or dashboard and don't want to add an MCP client to the mix, the same account, market, trading, risk, and macro capabilities are available as a plain HTTPS REST API, gated by the same subscription tiers.
Typical workflows
Pre-trade checklist in chat.
- Ask for current spread and recent bars.
- Ask for one or two indicators (for example RSI + ATR).
- Ask for position size by risk percentage.
- Ask for a dry run of the proposed order.
- If the preview is correct, explicitly request
dry_run: false.
Open risk review.
- Ask for open positions and account margin state.
- Ask for correlation/risk summary.
- Ask for concentration warning by symbol or direction.
- Ask for specific adjustments (partial close, breakeven, tighter SL) as dry runs first.
Post-trade review.
- Ask for the trade result with context (session, spread, volatility).
- Ask for one behavior insight and one process improvement.
- Save a trade note so the decision is documented.
What it doesn't do
MTContext does not make trading decisions for you and does not run an autonomous strategy loop. It gives an AI client the ability to read your terminal's state and, when explicitly instructed, act on it. The AI client — Claude, Cursor, whatever you've connected — is the one deciding what to ask for; MTContext is the bridge, not the strategy.
It also does not bypass broker constraints. Lot limits, margin requirements, and symbol trading rules still apply exactly as enforced by your broker and terminal.
How it's different
Most MT4/5-to-AI bridges require installing a local Python process and MCP server on the same machine as your terminal — workable for developers, not for most traders. MTContext's Expert Advisor connects outbound to a hosted cloud server, so the only thing you install in the terminal is a single .ex4/.ex5 file. Your AI client connects to the hosted MCP endpoint from anywhere — it doesn't need to be on the same machine, or even the same network, as your terminal.
Subscription tiers
Tools are gated by capability — read, analytics, or trade — and your subscription tier determines which capabilities are unlocked. Read-only tools (prices, positions, account state) are available broadly; trading tools require Pro tier or above. Check your dashboard for the exact breakdown for your plan. The same tiers and capabilities apply to the REST API.
Next step
Pick your client in Setup and connect. It takes about five minutes if you already have a terminal key.
Good operating habits
- Keep prompts specific: include symbol, timeframe, and intended risk.
- Prefer explicit confirmations before live actions.
- Use dry runs whenever you change symbol, account, or strategy template.
- Rotate terminal keys periodically and after any suspected leak.