Skip to main content
Glama

mcp-mt5

Model Context Protocol server for the MetaTrader 4/5 build pipeline. Compile MQL sources, deploy compiled EAs, run Strategy Tester, parse reports, tail logs — all driven by an LLM agent without touching the MetaTrader UI.

CI Python License: MIT Platform


What this is — and what it isn't

✅ This server

❌ Not this server

MetaTrader dev harness — compile, deploy, backtest, parse

Live trading (orders, positions, quotes)

Wraps MetaEditor64.exe / terminal64.exe CLI directly

Wraps the MetaTrader5 Python package

Runs entirely offline against installed terminal

Connects to a broker server

Iterates strategies before they go live

Executes strategies in production

Use case: an LLM agent edits .mq5 source → compiles → deploys → runs Strategy Tester → reads report → adjusts → repeats. No broker login, no human in the loop, no risk of real-money execution.

For runtime trading, pair this with a live-trading MCP — they target different layers and compose well.


Related MCP server: Meta-Trader-MCP

Tools

27 tools (plus 16 deprecated aliases until 0.6.0) and 2 resource templates. Every tool carries MCP annotations (read-only / destructive hints), a typed output schema and per-parameter descriptions; failures are reported as tool errors, never as a normal result with an error key.

🔍 Discovery & terminal selection

Tool

Description

env_info

Resolved paths, terminal hash, edition, and missing-component issues

list_terminals

Every MT4/5 terminal data folder under %APPDATA%\MetaQuotes\Terminal with its origin.txt install path

select_terminal

Switch the active terminal data folder mid-session by origin path, hash, or install dir

🔨 Build & deploy

Tool

Description

compile

MetaEditor CLI on a .mq4/.mq5/.mqh. ok requires a Result: line with 0 errors and a freshly written binary (MetaEditor's exit code is 1 even on success and is ignored). syntax_only=true uses /s for a fast check without a binary

compile_and_deploy

Compile, then copy the fresh .ex4/.ex5 into Experts/

deploy

Copy a compiled binary into Experts/ or a .mqh into Include/, chosen by extension

smoke_test

Compile + deploy + 1-day headless backtest + journal scan for runtime errors

list_experts

Enumerate Experts/ (defaults to *.ex5 on MT5, *.ex4 on MT4)

🔎 Source analysis & lint

Tool

Description

inspect_source

input declarations, #include tree with missing files, MetaEditor doc blocks, whole-word symbol search, duplicate magic numbers (aspects=[...])

analyze_mql

Structural lint (missing OnInit/OnTick, unused inputs, hardcoded magic/symbol), MT4-style deprecated calls with MT5 replacements, LOC/function/nesting metrics (checks=[...])

validate_tester_ini

Required keys, date format, numeric ranges; cross-checks [TesterInputs] against the EA source

✏️ Format & refactor

Tool

Description

format_mql

clang-format with MQL literals (D'…', C'…') and input group/#property lines protected. Dry run unless write=true; keeps the file's encoding

rename_symbol

Whole-word rename across a tree, dry_run by default

extract_function

Brace-aware extraction of a line range into a helper (inline or into a .mqh), dry_run by default

📊 Strategy Tester

Tool

Description

patch_tester_ini

Update Section.Key values in a tester.ini in place (encoding preserved)

gen_tester_inputs

Build a [TesterInputs] block from the EA's inputs, optionally written into an ini

run_backtest

Launch terminal64.exe /config:tester.ini, wait for ShutdownTerminal=1, stream progress notifications, return report_path, report_uri, tester log and journal notes (history_synchronized, final_balance, test_passed, start_time_changed…). Retries once when the tester started before the history download finished

start_backtest / get_backtest / cancel_backtest

Background runs: launch and get a run_id immediately, poll status (or list all runs), kill. Runs persist to .mt5tmp/runs/

read_tester_report

Parse an MT5/MT4 report into a ~50-field summary; response_format="detailed" adds every trade row. The HTML itself is served as the mt5://report/{id} resource

compare_reports

Diff two reports key by key with absolute/percent deltas; pass guards to get violations/ok

read_optimization

Read a Tester/cache/*.opt cache using the layout MetaQuotes published: header, optimised inputs, and the top N passes by any criterion over all passes

kill_terminal

Kill terminals launched by this server (all_instances=true for every instance)

📝 Logs & snapshots

Tool

Description

tail_log

Tail Files/LiveLog.txt, the daily Experts log MQL5/Logs/YYYYMMDD.log, the terminal Journal logs/YYYYMMDD.log, or the newest tester log; optionally parsed into {ts, source, message} (both the tab-separated and dated journal formats)

snapshot_sources / list_snapshots

Freeze source files into a timestamped folder with a manifest.json; list them

📡 MCP resources

URI

Description

mt5://log/{mode}

Last 500 lines of livelog, journal, terminal or tester

mt5://report/{id}

Full HTML of a report returned by read_tester_report / get_backtest (latest = newest on disk)

Deprecated aliases (removed in 0.6.0)

The pre-0.5.0 tool names still work in 0.5.x as thin wrappers whose descriptions start with DEPRECATED. Set MCP_MT5_LEGACY_TOOLS=0 to hide them and shrink the tool catalogue to the 27 tools above.

Deprecated name

Use instead

syntax_check

compile(syntax_only=true)

deploy_ea, install_include

deploy

extract_inputs, resolve_includes, extract_doc, find_symbol, find_magic_collision

inspect_source

lint_basic, check_deprecated, code_metrics

analyze_mql

format_check

format_mql (dry run by default)

parse_optimization, top_passes

read_optimization

regression_check

compare_reports(guards=…)

list_backtests

get_backtest() without run_id

mt5://livelog, mt5://journal, mt5://tester-log

mt5://log/{mode}


Quick start

Install

pip install mcp-mt5

Requires Windows + an installed MetaTrader 4 or 5 terminal.

Register with an MCP client

Most MCP clients accept a JSON entry under mcpServers. The server inherits its configuration from environment variables:

{
  "mcpServers": {
    "mt5": {
      "command": "mcp-mt5",
      "env": {
        "MT5_INSTALL": "C:\\Program Files\\MetaTrader 5"
      }
    }
  }
}

Refer to your client's documentation for the exact config file location.

Verify the install

Once registered, ask your agent to call env_info:

{
  "edition": "mt5",
  "install": "C:\\Program Files\\MetaTrader 5",
  "terminal_hash": "<32-char-hex-hash>",
  "metaeditor": "C:\\Program Files\\MetaTrader 5\\MetaEditor64.exe",
  "experts_dir": "C:\\Users\\<you>\\AppData\\Roaming\\MetaQuotes\\Terminal\\<hash>\\MQL5\\Experts",
  "issues": []
}

An empty issues array means everything is wired up correctly.


Configuration

Resolution priority for the MetaTrader install + data folder:

  1. Explicit env vars (below)

  2. Auto-scan of %APPDATA%\MetaQuotes\Terminal\*\origin.txt for a folder whose origin matches MT5_INSTALL

  3. Portable mode fallback (data colocated with install dir)

Env var

Default

Notes

MT5_INSTALL

C:\Program Files\MetaTrader 5

Install dir containing terminal64.exe

MT5_DATA

(auto-detected)

%APPDATA%\MetaQuotes\Terminal\<hash>

MT5_TERMINAL_HASH

(auto-detected)

32-char folder name

MT5_EDITION

mt5

Set to mt4 for MetaTrader 4

Running the server from WSL

The server can run inside WSL and drive the Windows MetaTrader install through WSL interop. The install (/mnt/c/Program Files/MetaTrader 5) and the data folder (scanned under /mnt/*/Users/*/AppData/Roaming/MetaQuotes/Terminal, matched through origin.txt) are auto-detected; MT5_INSTALL / MT5_DATA override them. Paths handed to MetaEditor/terminal are translated with wslpath -w automatically. Interop must be enabled (/etc/wsl.conf [interop] enabled=true, then wsl --shutdown).

MT4 support

Set MT5_EDITION=mt4 and point MT5_INSTALL at your MT4 install. The server switches to metaeditor.exe (32-bit), terminal.exe, and the MQL4/ data tree automatically.


Example workflow

A typical LLM-driven iteration loop:

1. env_info                                          → verify paths
2. compile_and_deploy source="C:\\...\\MyEA.mq5"       → 0 errors, .ex5 deployed ✅
3. patch_tester_ini config="tester.ini" updates={
     "Tester.Symbol": "EURUSD",
     "Tester.FromDate": "2025.01.01",
     "TesterInputs.RiskPct": "1.5"
   }
4. run_backtest config="tester.ini" wait=true
5. read_tester_report path=<report_path from step 4>  → summary.net_profit = 1234.56
                                                       summary.profit_factor = 1.45
6. tail_log mode="tester" lines=200 structured=true  → diagnose journal warnings
7. <edit Signal.mqh based on findings>
8. → loop back to step 2

A sample tester.ini

; Launch: terminal64.exe /config:tester.ini
; Period codes: M1=1, M5=5, M15=15, H1=16385, H4=16388, D1=16408
; Model: 0=Every tick, 1=1 min OHLC, 4=Real ticks

[Tester]
Expert=MyEA
Symbol=EURUSD
Period=M15
Model=1
FromDate=2024.01.01
ToDate=2024.12.31
Deposit=10000
Currency=USD
Leverage=500
Visual=0
ShutdownTerminal=1     ; required so run_backtest can wait for the run to finish
Report=tester_report

[TesterInputs]
; ParamName=value||start||step||stop||(N=fixed|Y=optimize)
; RiskPct=1.0||0.1||0.1||3.0||N

A more complete sample lives at examples/tester.ini.


Development

git clone https://github.com/PHUICMT/mcp-mt5
cd mcp-mt5
pip install -e ".[dev]"
pytest                    # runs the test suite (no MetaTrader needed)
ruff check src tests      # lints

CI runs on Windows for Python 3.10, 3.11, and 3.12 against every push to main. Tagging a release (e.g. v0.2.0) triggers an OIDC publish to PyPI.

Project layout

mcp-mt5/
├── src/mcp_mt5/
│   ├── server.py        # FastMCP tool definitions
│   ├── paths.py         # Layout detection + origin.txt scan
│   ├── parsers.py       # Compile log, tester report/journal parsers, encoding helpers
│   ├── analysis.py, lint.py, formatting.py, refactor.py, ast_refactor.py
│   ├── optimization.py  # .opt cache reader (documented TesterOptCache layout)
│   ├── reports.py, snapshot.py, smoke.py, workdir.py
├── tests/               # pytest suite, no live MT5 required
├── examples/            # Sample tester.ini + client config
└── .github/workflows/   # CI + PyPI release

Limitations

  • Windows-only. MetaTrader CLI binaries don't ship for Linux/macOS. Wine ports may work but are untested.

  • No live broker access. This server intentionally never authenticates to a broker. Use a separate MCP server for runtime trading.

  • Tester report parsing is best-effort. MetaTrader's HTML output isn't a stable schema; the raw HTML is also returned alongside the parsed structure so you can fall back to text inspection when needed.

  • Optimisation caches (Tester/cache/*.opt) are parsed with the layout MetaQuotes published in MQL5 Programming for Traders. Files with an unrecognised record layout are reported as such rather than guessed.


Roadmap

All v0.3.x roadmap items shipped in v0.4.0. Future ideas:

  • Real tree-sitter MQL grammar for extract_function (current implementation is brace-counting + regex)

  • WebSocket transport for long-lived sessions (currently stdio only)

  • Linux/Wine port for non-Windows agents


License

MIT © 2026 PHUICMT

Available Tools

43 tools
analyze_mqlA
Read-onlyIdempotent

Static checks on MQL source without compiling: structural lint (missing OnInit/OnTick, unused inputs, hardcoded magic/symbol), MT4-style deprecated API calls with MT5 replacements, and size/nesting metrics.

Replaces lint_basic / check_deprecated / code_metrics. Use compile(syntax_only=true) for real compiler diagnostics.

Inputs: root, checks, source. Returns: object

cancel_backtestA
Destructive

Kill the terminal of a running background backtest. No report is produced for a cancelled run.

Inputs: run_id (required). Returns: object

check_deprecatedA
Read-onlyIdempotent

DEPRECATED (removed in 0.6.0): use analyze_mql(source, checks=["deprecated"]).

Inputs: source (required). Returns: object

code_metricsA
Read-onlyIdempotent

DEPRECATED (removed in 0.6.0): use analyze_mql(source or root, checks=["metrics"]).

Inputs: root, source. Returns: object

compare_reportsA
Read-onlyIdempotent

Diff two tester reports key by key (absolute and percent deltas), optionally checking guard thresholds.

With guards, a profit-style metric below its threshold or a drawdown/loss metric above it is a violation. Accepting only improvements across many iterations is selection bias; treat guards as a sanity gate, not proof of robustness. Replaces regression_check.

Inputs: baseline (required), candidate (required), guards. Returns: object

compileA

Compile one MQL source with MetaEditor and return structured diagnostics.

ok is true only when the log has a Result: line with 0 errors AND (unless syntax_only) a fresh .ex5/.ex4 was produced (binary_fresh). Fix every entry in errors (file, line, col, code, message) and call again; warnings do not block deployment. Takes 1-30 s. To also copy the binary into Experts/ use compile_and_deploy.

Inputs: source (required), include, log_file, syntax_only, +1 more. Returns: object

compile_and_deployA
Destructive

Compile a source and, if it succeeds, copy the fresh binary into the terminal's Experts/ folder.

Use this before run_backtest or smoke_test. Returns the compile result plus deploy; ok is false if either stage failed (compile errors are in compile.errors).

Inputs: source (required), ea_name. Returns: object

deployA
Destructive

Copy a compiled EA into the terminal's Experts/ folder or a .mqh header into Include/, chosen by extension.

Overwrites an existing file of the same name. Use compile_and_deploy to build and copy in one step.

Inputs: file (required), name. Returns: object

deploy_eaA
Destructive

DEPRECATED (removed in 0.6.0): use deploy(file, name).

Inputs: source_ex (required), name. Returns: object

env_infoA
Read-onlyIdempotent

Resolve and report MT4/5 paths, terminal hash, and missing-component issues.

Inputs: no named parameters. Returns: object

extract_docA
Read-onlyIdempotent

DEPRECATED (removed in 0.6.0): use inspect_source(source, aspects=["docs"]).

Inputs: source (required). Returns: object

extract_functionA
Destructive

Extract a contiguous block of lines into a new helper function.

Brace-counting + regex param detection — not a full AST parser. Returns the proposed helper, call site, and parameter list. Set dry_run=False to write.

Inputs: source (required), line_end (required), new_name (required), line_start (required), +3 more. Returns: object

extract_inputsA
Read-onlyIdempotent

DEPRECATED (removed in 0.6.0): use inspect_source(source, aspects=["inputs"]).

Inputs: source (required). Returns: object

find_magic_collisionA
Read-onlyIdempotent

DEPRECATED (removed in 0.6.0): use inspect_source(root=..., aspects=["magic"]).

Inputs: root (required), var_pattern. Returns: object

find_symbolA
Read-onlyIdempotent

DEPRECATED (removed in 0.6.0): use inspect_source(root=..., symbol=...).

Inputs: root (required), symbol (required), exts, limit. Returns: object

format_checkA
Read-onlyIdempotent

DEPRECATED (removed in 0.6.0): use format_mql(source) (dry run by default).

Inputs: source (required), style. Returns: object

format_mqlA
Destructive

Format an MQL file with clang-format (MQL literals such as D'…' and input group lines are protected).

Default is a dry run reporting whether the file would change; pass write=true to apply. Replaces format_check.

Inputs: source (required), style, write. Returns: object

gen_tester_inputsA
Destructive

Generate a [TesterInputs] block from EA inputs.

If write_to points at a tester.ini, the block is appended/replaced in-place.

Inputs: source (required), write_to. Returns: object

get_backtestA
Read-onlyIdempotent

Status of a background backtest (running / completed / timeout / cancelled) with report path and journal notes when done; without run_id, lists all runs newest first.

Inputs: run_id, tail_lines. Returns: object

inspect_sourceA
Read-onlyIdempotent

Read-only facts about MQL source: input declarations, #include tree (with missing files), MetaEditor doc blocks, symbol usages, duplicate magic numbers.

Replaces extract_inputs / resolve_includes / extract_doc / find_symbol / find_magic_collision.

Inputs: root, limit, source, symbol, +3 more. Returns: object

install_includeA
Destructive

DEPRECATED (removed in 0.6.0): use deploy(file, name) with a .mqh file.

Inputs: source (required), target_name. Returns: object

kill_terminalA
Destructive

Force-kill terminal processes launched by this server (run_backtest / smoke_test).

By default only PIDs this server started are killed, so a live-trading terminal on the same machine is never touched. Pass all_instances=True to taskkill every terminal64.exe/terminal.exe of the configured edition (destructive).

Inputs: all_instances. Returns: object

lint_basicA
Read-onlyIdempotent

DEPRECATED (removed in 0.6.0): use analyze_mql(source, checks=["lint"]).

Inputs: source (required). Returns: object

list_backtestsA
Read-onlyIdempotent

DEPRECATED (removed in 0.6.0): use get_backtest() without run_id.

Inputs: no named parameters. Returns: object

list_expertsA
Read-onlyIdempotent

List compiled EAs in Experts/ (defaults to *.ex5 for MT5, *.ex4 for MT4).

Inputs: pattern, recurse. Returns: object

list_snapshotsA
Read-onlyIdempotent

List all snapshot folders under dest.

Inputs: dest (required). Returns: object

list_terminalsA
Read-onlyIdempotent

Enumerate all MetaTrader terminal data folders under %APPDATA%\MetaQuotes\Terminal.

Inputs: no named parameters. Returns: object

parse_optimizationD
Read-onlyIdempotent

DEPRECATED (removed in 0.6.0): use read_optimization(...).

Inputs: path, expert, period, sample, +1 more. Returns: object

patch_tester_iniA
Destructive

Set keys in a tester.ini in place, keeping the file's encoding and other lines.

updates maps "Section.Key" to a value, e.g. {"Tester.Symbol": "EURUSD", "Tester.FromDate": "2025.01.01", "TesterInputs.RiskPct": "1.5"}. Missing sections/keys are added. Run validate_tester_ini afterwards; always set Deposit, Currency, Leverage, Model, Optimization, Visual, UseLocal/UseRemote/UseCloud and Report explicitly so a run does not inherit the machine's last UI state.

Inputs: config (required), updates (required). Returns: object

read_optimizationA
Read-onlyIdempotent

Read an MT5 optimisation cache (.opt): header, optimised inputs, pass count and the top N passes by criterion.

Ranking uses every pass in the cache. The best pass of a genetic run is a biased sample; prefer robust neighbourhoods over the single top row. Replaces parse_optimization / top_passes.

Inputs: path, top_n, expert, period, +3 more. Returns: object

read_tester_reportA
Read-onlyIdempotent

Parse a Strategy Tester HTML report (MT5 or MT4) into summary metrics; the full HTML stays behind report_uri.

Prefer passing the report_path returned by run_backtest/get_backtest; without path the newest report on disk is used, which may belong to an older run. summary values are the strings printed in the report ("10 000.00", "359.61 (3.34%)"); use compare_reports / regression_check for numeric comparison. Read the mt5://report/{id} resource only when you need the original HTML.

Inputs: path, max_trades, raw_truncate, response_format. Returns: object

regression_checkA
Read-onlyIdempotent

DEPRECATED (removed in 0.6.0): use compare_reports(baseline, candidate, guards=...).

Inputs: baseline (required), candidate (required), guards. Returns: object

rename_symbolA
Destructive

Rename a symbol across MQL files (whole-word match). dry_run=True previews only.

Inputs: new (required), old (required), root (required), dry_run. Returns: object

resolve_includesC
Read-onlyIdempotent

DEPRECATED (removed in 0.6.0): use inspect_source(source, aspects=["includes"]).

Inputs: source (required), mql_root. Returns: object

run_backtestA

Run a Strategy Tester backtest headlessly from a tester.ini and wait for it to finish.

Requires the EA to be deployed first (compile_and_deploy / deploy_ea) and ShutdownTerminal=1 in the ini, otherwise the call runs until timeout_sec. Typically 1-30 minutes; progress notifications are sent every few seconds while waiting. If the tester reports "cannot synchronize history" (first headless start after login), the run is retried once automatically. For long runs prefer start_backtest + get_backtest. Returns run_id, report_path (resolved from Report= in the ini), latest_tester_log and journal_notes/warnings; if warnings mentions start_time_changed the tester moved FromDate because history was missing, so the report is not comparable with other runs.

Inputs: config (required), wait, portable, timeout_sec. Returns: object

select_terminalA

Switch the active terminal data folder for this session.

Provide one of: origin (install path stored in origin.txt), hash (32-char folder name), or install (auto-scan for the matching origin).

Subsequent tool calls will use the new layout until the server restarts.

Inputs: hash, origin, edition, install. Returns: object

smoke_testA
Destructive

One-call health check for an EA: compile, deploy, run a 1-day headless backtest, scan the journal.

Catches problems that compile cleanly but fail at runtime (OnInit errors, array out of range, divide by zero). ok is true only if every stage passes; stage tells where it stopped. Overwrites the EA of the same name in Experts/. Takes 1-10 minutes.

Inputs: source (required), days, period, symbol, +2 more. Returns: object

snapshot_sourcesA
Idempotent

Freeze a copy of source files into a timestamped folder under dest.

Inputs: dest (required), sources (required), label. Returns: object

start_backtestA

Launch a backtest in the background and return immediately with a run_id.

Poll get_backtest(run_id) until status is no longer "running"; the final record carries report_path and journal_notes. Runs are also written to .mt5tmp/runs/<run_id>.json next to the ini. Only one terminal per install can test at a time.

Inputs: config (required), portable, timeout_sec. Returns: object

syntax_checkA

DEPRECATED (removed in 0.6.0): use compile(source, syntax_only=true).

Inputs: source (required), timeout_sec. Returns: object

tail_logA
Read-onlyIdempotent

Tail a MetaTrader log: the EA's LiveLog.txt, the daily Experts journal, or the latest tester journal.

Use mode="tester" right after a backtest to see runtime errors and the effective test period.

Inputs: date, mode, lines, structured. Returns: object

top_passesA
Read-onlyIdempotent

DEPRECATED (removed in 0.6.0): use read_optimization(criterion=..., top_n=...).

Inputs: n, expert, period, symbol, +3 more. Returns: object

validate_tester_iniA
Read-onlyIdempotent

Sanity-check a tester.ini. If source given, cross-check inputs vs EA declarations.

Inputs: config (required), source. Returns: object

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    B3.2/5.0

    Scored across 43 tools

    Disambiguation3/5

    The active tools are mostly distinct, but the 16 deprecated aliases duplicate the same operations as active tools (e.g., extract_inputs vs inspect_source, parse_optimization vs read_optimization), creating confusion. Some overlap also exists between compile/compile_and_deploy/deploy and run_backtest/start_backtest/smoke_test, though descriptions help clarify boundaries.

    Naming Consistency4/5

    Most tool names follow a clear verb_noun pattern (list_snapshots, run_backtest, read_tester_report, validate_tester_ini). Minor deviations like env_info, compile_and_deploy, and smoke_test are understandable exceptions, and deprecated names also generally follow similar conventions.

    Tool Count2/5

    With 43 tools, the surface is far too large, even though 16 are deprecated. The active count of 27 is still heavy, and several active tools are composites (compile_and_deploy, smoke_test) that could be consolidated without losing capability.

    Completeness4/5

    The server covers the major MT5 automation lifecycle: terminal selection, compile/deploy, backtest launch and monitoring, report parsing, source inspection, static analysis, formatting, optimization reading, and snapshots. Minor gaps exist (no explicit tester.ini creation tool, no undeploy/removal tool), but workflows do not hit dead ends.

    Maintenance

    ActivityStale
    ResponsivenessOutdated data

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      Not graded
      quality
      D
      maintenance
      A standalone Model Context Protocol (MCP) server that enables AI assistants to interact with the cTrader trading platform.
      15
      -
    • A
      license
      Not graded
      quality
      A
      maintenance
      Exposes a unified AI interface to MetaTrader 5 over the Model Context Protocol, enabling live quotes, historical data, technical indicators, order execution, position management, and headless backtests.
      134 PyPI
      MIT
    • A
      license
      Not graded
      quality
      C
      maintenance
      A local-first MCP server that bridges AI coding agents with MetaTrader 5 for inspection, market data, MQL5 development, compiling, Strategy Tester review, workspace sync, logs, audit trails, demo trading, and carefully gated live trading.
      MIT
    • F
      license
      Not graded
      quality
      D
      maintenance
      Universal MCP server that enables AI assistants to control MetaTrader 5 directly, executing trades, managing data, and automating MT5 operations through natural language.
      4
      -