Finance & Investment Banking metatrader-5mcp-servertrading-automationforexorder-managementmarket-datawebsocket-streaming

MetaTrader 5 Trading Terminal Assistant

Connect AI assistants to MetaTrader 5 via natural language so account checks, quotes, orders and position closes become one-line commands over real money.

FollowSkills review · FSRS-2.0
Use with care
47/ 100 5-point scale 2.4 / 5
1 2 3 4 5 6
1Trust11 / 25 · 2.2/5

The skill transparently lists all tools and parameters; the README includes a risk disclaimer and SSE security warnings, and credentials can be supplied via .env. However, SKILL.md explicitly instructs 'execute immediately — no additional confirmation needed' for real-money trades, abandoning user confirmation, and batch-destructive operations (close_all_positions etc.) lack confirmation or recovery paths — the main deduction. No malware or exfiltration red lines found.

2Reliability9 / 20 · 2.3/5

SKILL.md is internally self-consistent: the two-step market-order-with-SL/TP workflow, error classification with remediation suggestions, and empty-CSV handling are documented. But it depends on an external metatrader MCP server, tests require live MT5 credentials (skipped otherwise), and key trading paths cannot be reproduced offline with no statically verifiable failure-mode control — scored mid-low.

3Adaptability8 / 15 · 2.7/5

Trigger description is clear (trading, MT5, forex keywords), scenarios and some boundaries are defined (symbol format variance, demo vs live hinted only via account_type). But it fully depends on the MetaTrader 5 terminal (Windows-only) and a broker account, declares no Chinese-language support, and overly broad trigger words risk false invocation on stock-analysis queries — deducted.

4Convention9 / 15 · 3.0/5

Docs are well-layered (tool tables, workflows, formatting, domain knowledge, error handling); MIT license explicit, version 0.5.1, maintainer identity and PyPI publication traceable. But the skill has no version/changelog, tests requiring a live account limit maintainability verification, and README-listed tools (get_candles_by_date, get_symbol_info) are missing from the SKILL.md tool table — sync risk, deducted.

5Effectiveness6 / 15 · 2.0/5

Goals are clear, output templates are directly usable, and the two-step SL/TP workflow adds real value over unstructured natural-language trading. But static review cannot verify execution correctness, the cost/benefit of batch-close operations is unproven, and templates are examples only — deducted.

6Verifiability4 / 10 · 2.0/5

A test file exists (tests/metatrader_client/test_account.py) but covers only the account module of the client library and requires live credentials; the skill's claimed trade-execution key paths are untested. No third-party execution evidence, no CI test workflow (only Claude code-review workflows) — conclusions are not independently reproducible, deducted.

Evidence confidence:Low Reviewed Sep 10, 2026 Reviewed revision bd2ff591e32c
Before you use it
  • SKILL.md instructs the AI to execute real-money trades immediately without confirmation; users should add their own confirmation layer or connect only a demo account.
  • Batch operations (close_all_positions, cancel_all_pending_orders) have no undo/recovery path; false invocation could cause real financial loss.
  • Functionality fully depends on a Windows MetaTrader 5 terminal and broker credentials; mainland-China users must verify broker/MT5 reachability, and no Chinese support is declared.
  • Static review executed no code; the test suite requires live account credentials, so key trading paths lack third-party reproducible verification.
  • Some tools claimed in the README are absent from the skill's tool table, indicating possible doc/implementation drift.
Review evidence [1][2][3][4][5][6][7][8][9]
See the full review method →

What does this skill do, and when should you use it?

MetaTrader MCP Server is a bridge that connects AI assistants like Claude and ChatGPT to the MetaTrader 5 platform, and it ships with a pre-built Trading Terminal Assistant skill. It exposes 32 trading tools over the MCP protocol covering account queries, market data, market and pending orders, position management, and trade history. The skill gives the AI tool parameter specs, output formatting templates, and MT5 domain knowledge so trades execute precisely. Because it operates real money, the repository carries an explicit risk disclaimer.

Calls the metatrader MCP server (tools prefixed mcp__metatrader__) to: query balance, equity, margin and leverage; fetch live symbol prices and candle data; place market and pending orders (with a two-step workflow for SL/TP); modify stop loss and take profit on positions; bulk-close positions by symbol or profit/loss state; cancel pending orders; query deal and order history by date range; and render results as terminal-style tables.

  1. A day trader who wants to ask Claude for account equity and place orders by voice or text instead of clicking through the MT5 terminal
  2. A quant developer wiring MT5 into their own scripts via the REST API endpoints like /api/v1/order/market
  3. A monitoring setup that streams live bid/ask/spread ticks from the WebSocket quote server into a dashboard
  4. A multi-position holder who wants the AI to list all positions and selectively close the profitable or losing ones
  5. An Open WebUI user who wants trading tools available inside their own chat interface

What are this skill's strengths and limitations?

Pros
  • Full MT5 trading lifecycle coverage: account, market data, order placement, modification, closure, cancellation, and history across 32 tools
  • Four integration surfaces: MCP server, REST API, WebSocket quote stream, and a Python client library
  • The bundled skill includes tool tables, workflows, output formats, and error-handling guidance that reduce hallucination and parameter errors
  • Credentials stay on your machine via .env or CLI flags; published on PyPI for easy install
Limitations
  • Operates real money and the skill instructs skipping confirmation, raising the risk of accidental trades
  • The MetaTrader 5 Python package only works with a Windows terminal; macOS/Linux users must run the SSE server on a Windows VPS
  • SSE remote mode has no authentication — you must add a firewall, authenticated reverse proxy, or SSH tunnel yourself
  • place_market_order lacks SL/TP parameters, forcing a two-step workflow with a window where the position sits unprotected
  • README offers no evidence of test coverage or independent security audits

How do you install this skill?

Prerequisites: Python 3.10+, the MetaTrader 5 terminal, and MT5 account credentials (login, password, server name), with algorithmic trading enabled in MT5 (Tools → Options → Expert Advisors → Allow algorithmic trading). Install the package: pip install metatrader-mcp-server. Install the skill by copying or symlinking claude-skill/trading into your skills directory. Claude Code: mkdir -p .claude && ln -s ../claude-skill .claude/skills (or copy to .claude/skills/trading). Claude Desktop: copy to ~/Library/Application Support/Claude/skills/trading (macOS) or %APPDATA%\Claude\skills\trading (Windows).

How do you use this skill?

Register the metatrader MCP server in your Claude Desktop config first (stdio mode needs --login, --password, --server, --transport stdio). Once the skill is installed, invoke with /trading or just ask naturally, e.g. 'Show me my account dashboard', 'Buy 0.1 lots of EURUSD with SL at 1.0800', 'Close all profitable positions'. Market orders do not take SL/TP parameters — place the order, then call modify_position. The skill instructs immediate execution on trade requests without a second confirmation.

Related skills