Connect Claude, Cursor and VS Code to MetaTrader 5 with MCP

fxapis runs a hosted Model Context Protocol (MCP) server for its MetaTrader 5 API. Point an AI assistant at it with an fxapis API key, and it can list your accounts, read positions and history, bring accounts online, and — only if the key allows it — place, close and modify trades. fxapis-mcp-examples has a tested config for each client; this post walks through setting one up safely.

How it works

• The server: https://api.fxapis.com/mcp, over Streamable HTTP. It is stateless, so any client that supports remote MCP servers with a custom header can use it.
• Authentication: your fxapis API key, as Authorization: Bearer fx_live_…. OAuth sign-in is not available yet.
• The same rules as the REST API. Each tool calls the ordinary API with your key. Scopes, plan limits, idempotency and the workspace kill switch apply exactly as they do to your own code — the MCP server cannot do anything the key could not do directly.
• Deliberately missing: connecting an MT5 account and managing API keys. Connecting takes the broker password, and a broker credential should never pass through a language model's context. Do both in the console.

Create a key just for the agent

The key's scopes are the control that actually holds, whatever the model decides. Give the assistant the fewest it needs:

A tool the key does not cover answers MISSING_SCOPE, and the agent is told to ask you rather than work around it. Revoke the key and the agent is cut off on its next call. API keys vs broker logins explains why the key, not the MT5 password, is what you hand out.

Connect your client

Claude Code, from a terminal:

claude mcp add --transport http fxapis https://api.fxapis.com/mcp \
  --header "Authorization: Bearer $FXAPIS_KEY"

Cursor, in ~/.cursor/mcp.json or the project's .cursor/mcp.json:

{
  "mcpServers": {
    "fxapis": {
      "url": "https://api.fxapis.com/mcp",
      "headers": { "Authorization": "Bearer ${env:FXAPIS_KEY}" }
    }
  }
}

VS Code (GitHub Copilot agent mode), in .vscode/mcp.json — the key is asked for once and stored securely, never written to the file:

{
  "inputs": [{ "type": "promptString", "id": "fxapis-key", "description": "fxapis API key", "password": true }],
  "servers": {
    "fxapis": {
      "type": "http",
      "url": "https://api.fxapis.com/mcp",
      "headers": { "Authorization": "Bearer ${input:fxapis-key}" }
    }
  }
}

The repository also has Claude (claude.ai and Claude Desktop), the OpenAI Responses API and the Anthropic Messages API, plus plain Python and Node scripts that call the tools with no model involved. To check a key from a terminal, list the tools with the MCP Inspector:

npx @modelcontextprotocol/inspector --cli https://api.fxapis.com/mcp --transport http \
  --header "Authorization: Bearer $FXAPIS_KEY" --method tools/list

Keeping an agent safe around real money

An agent can misread a request, act on text planted in something it was asked to summarize, or loop. The protections, strongest first:

1. The key's scopes, enforced on every call.
2. Demo accounts while you learn how an agent behaves — fxapis has no simulated market.
3. Your plan's limits, which refuse opening orders past the allowance; closing is never refused for billing.
4. Your client's approvals. Trading tools are marked destructive, so Claude Code and Cursor ask before a trade. Do not choose "always allow" for trading tools on a live account.

Every order an agent places is recorded like any other, with the key that placed it.

Things to try

• "Which of my MT5 accounts are online, and what positions are open on each?"
• "How much margin would 0.5 lots of XAUUSD need on account 2?"
• "Move the stop loss on my gold position to break-even." (needs trading:reduce)
• "Did anything go wrong with today's orders? Show me any that are still unknown."

Where to go next

• The MCP guide has the full tool list and the scope each tool needs.
• The MCP server on the website, and SDKs and examples for building your own agent on the API directly.