(Thought) Autonomous AI "Research Agent" for CommunityPower EA (MT5 + MCP)

The Goal

is to allow an LLM (like claude sonnet 4.5 / gemini3.0 or cli via Cursor/Antigravity/kilocode/other IDE) to:

1. Read our complex .set files intelligibly.
2. Understand the parameters by cross-referencing the official docs for CommunityPowerEA.
3. Modify the strategy safely based on natural language (e.g., "Make this set news-neutral"). 
4. (Future Phase) Run the backtest automatically and learn from the results with windows MCP. https://github.com/CursorTouch/Windows-MCP

The Challenge: Parsing .set Files

We all know the CommunityPower .set files use the custom MT5 optimization syntax: Key=Value||Start||Step||Stop||Enabled

Example Raw Line: Lot_Per_1000=0.01||0.01||0.001000||0.100000||N

If you feed this raw text to an AI, it gets confused by the optimization parameters (||...). It often hallucinates or breaks the file format when trying to edit it.

The Solution: A Custom MCP Parser To develop a Python-based MCP server that parses this syntax into a clean JSON structure that the AI can understand.

The Knowledge Gap (RAG Integration)

The EA uses numeric IDs for settings, which is hard for an AI to interpret.

• In File: Oscillators_Indicator=3
• To AI: "Set Indicator to 3" (The AI doesn't know what 3 is).

Such integrating the documentation so the MCP Server presents this to the AI as:

• To AI: Oscillators_Indicator=3 (Stochastic)

E.g:

def parse_cp_line(line):
    """
    Converts: Lot_Per_1000=0.01||0.01||0.001||0.1||N
    To: {'value': 0.01, 'opt_start': 0.01, 'opt_step': 0.001, ...}
    """
    if "=" not in line or "||" not in line:
        return None
        
    key, properties = line.split("=", 1)
    parts = properties.split("||")
    
    # The AI primarily cares about the 'active_value'
    return {
        "parameter": key.strip(),
        "active_value": parts[0], 
        "optimization_config": {
            "start": parts[1],
            "step": parts[2],
            "stop": parts[3],
            "enabled": parts[4]
        }
    }