neural-tradermcp - SKILL.md Agent Skill

name: "@neural-trader/mcp" description: "MCP server exposing 87+ neural trading tools for AI agent integration. Use when connecting trading capabilities to Claude or other AI agents via MCP, exposing market data and strategy tools through Model Context Protocol, or building agent-driven trading automation."

@neural-trader/mcp

Model Context Protocol (MCP) server wrapping the Neural Trader engine, exposing 87+ trading tools for seamless integration with Claude, AI agents, and MCP-compatible clients.

Quick Reference

Task	Code
Install	`npx @neural-trader/mcp@latest`
Import	`import { NeuralTraderMCP } from '@neural-trader/mcp';`
Create	`const mcp = new NeuralTraderMCP({ port: 3001 });`
Start	`await mcp.start();`
Stop	`await mcp.stop();`
Status	`const info = mcp.getStatus();`

Installation

Hub install (recommended): npx neural-trader@latest includes this package. Standalone: npx @neural-trader/mcp@latest See Installation Guide for the full ecosystem.

Key API

NeuralTraderMCP

The MCP server class exposing trading tools.

import { NeuralTraderMCP } from '@neural-trader/mcp';

const mcp = new NeuralTraderMCP({
  port: 3001,
  transport: 'stdio',
  engine: { mode: 'paper', capital: 100000 },
});

Constructor Options:

Option	Type	Default	Description
`port`	`number`	`3001`	Server port (for SSE transport)
`transport`	`string`	`'stdio'`	Transport: `'stdio'`, `'sse'`, `'ws'`
`engine`	`EngineOptions`	defaults	Neural Trader engine config
`authToken`	`string`	`undefined`	Optional authentication token
`enabledTools`	`string[]`	all	Subset of tools to enable
`readOnly`	`boolean`	`false`	Disable trade execution tools
`maxConcurrent`	`number`	`10`	Max concurrent tool calls

Methods:

Method	Returns	Description
`start()`	`Promise<void>`	Start the MCP server
`stop()`	`Promise<void>`	Stop the MCP server
`getStatus()`	`ServerStatus`	Server and engine status
`getTools()`	`ToolDefinition[]`	List all available tools
`enableTool(name)`	`void`	Enable a specific tool
`disableTool(name)`	`void`	Disable a specific tool

Tool Categories

The 87+ tools are organized into categories:

Category	Count	Description
Market Data	15	Price quotes, OHLCV, order books
Technical Analysis	20	RSI, MACD, Bollinger, SMA, EMA, etc.
Strategy Management	12	Create, configure, backtest strategies
Order Execution	10	Place, modify, cancel orders
Portfolio	8	Positions, balance, allocation
Risk Management	7	VaR, exposure, correlation, limits
Neural Models	8	Train, predict, evaluate alpha models
Backtesting	5	Run, analyze, compare backtests
Alerts	2	Price and condition alerts

Key MCP Tools

Market Data Tools:

Tool	Description
`get_quote`	Get real-time price quote for a symbol
`get_ohlcv`	Get OHLCV bars for a symbol and timeframe
`get_order_book`	Get current order book (L2 data)
`get_market_status`	Check if markets are open
`search_symbols`	Search for symbols by name

Strategy Tools:

Tool	Description
`create_strategy`	Create a new trading strategy
`backtest_strategy`	Backtest a strategy on historical data
`list_strategies`	List all configured strategies
`get_strategy_metrics`	Get strategy performance metrics
`optimize_strategy`	Optimize strategy parameters

Order Tools:

Tool	Description
`place_order`	Place a buy/sell order
`cancel_order`	Cancel a pending order
`get_orders`	List open/recent orders
`modify_order`	Modify order price/quantity

Portfolio Tools:

Tool	Description
`get_portfolio`	Get current portfolio summary
`get_positions`	List open positions
`get_balance`	Get account balance
`get_pnl`	Get realized/unrealized P&L

Common Patterns

Claude Desktop Integration

Add to Claude Desktop claude_desktop_config.json:

{
  "mcpServers": {
    "neural-trader": {
      "command": "npx",
      "args": ["@neural-trader/mcp@latest"],
      "env": {
        "NEURAL_TRADER_MODE": "paper"
      }
    }
  }
}

Programmatic MCP Server

import { NeuralTraderMCP } from '@neural-trader/mcp';

const mcp = new NeuralTraderMCP({
  transport: 'stdio',
  engine: { mode: 'paper', capital: 50000 },
  readOnly: false,
});

await mcp.start();

Read-Only Market Data Server

import { NeuralTraderMCP } from '@neural-trader/mcp';

const mcp = new NeuralTraderMCP({
  transport: 'sse',
  port: 3001,
  readOnly: true,
  enabledTools: ['get_quote', 'get_ohlcv', 'search_symbols', 'get_market_status'],
});

await mcp.start();

RAN DDD Context

Bounded Context: RANO Optimization

References

API reference: See references/commands.md
Full README
npm