⚙️

FLOX Trading Framework MCP

v0.6.6

Model DeductionLocal

flox-mcp from FLOX-Foundation/flox, run locally after `pip install flox-mcp`. Free and open source; live/paper trading requires a separate FLOX engine. run_backtest is an MVP sandbox — do not use untrusted code in production.

📖 Docs 🔗 Source

§ Quickstart

Add the following to your MCP client configuration (Cursor, Claude Desktop, etc.):

pip install flox-mcp && flox-mcp init writes ./.mcp.json; restart your MCP client.

Install:

pip install "flox-mcp[flox]"

MCP Client Configuration

{
  "mcpServers": {
    "flox": {
      "command": "flox-mcp"
    }
  }
}

Config file locations

Cursor: Settings → MCP, or edit ~/.cursor/mcp.json
Claude Desktop: edit claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/, Windows: %APPDATA%\Claude\)

§ Tools Reference

4 tools

run_backtest

Run Python strategy backtest on CSV dataset

Parameter	Type	Required	Description
strategy_code	string	✓	Strategy subclass source
dataset_path	string	✓	Absolute CSV path

compute_indicator

Compute FLOX indicator on float series

Parameter	Type	Required	Description
indicator	string	✓	Indicator name

scaffold_strategy

Generate strategy scaffold code

Parameter	Type	Required	Description
kind	string	✓	bar/trade/hybrid

docs_search

Search FLOX documentation

Parameter	Type	Required	Description
query	string	✓	Search query

§ Examples

Once configured, you can ask the AI directly:

CSV backtest

Use run_backtest to test a simple MA strategy on my BTC CSV

Local CSV backtest