NinjaTrader NinjaScript Knowledge Base

A comprehensive, LLM-ready knowledge base of the NinjaTrader Desktop SDK documentation for building indicators and strategies with AI coding assistants.

What’s Included

Total: 1170 documentation pages

Quick Start

Option 1: MCP Server (Recommended)

Use the MCP server for dynamic, on-demand access to the documentation:

{
  "mcpServers": {
    "ninjatrader": {
      "command": "npx",
      "args": ["-y", "ninjatrader-ninjascript-mcp"]
    }
  }
}

Add this to your MCP config file:

The MCP server provides tools to search docs, get specific pages, and follow recommended workflows.

Option 2: Static Files

Claude Code / Kiro

Copy CLAUDE.md to your project root, or add this to your existing CLAUDE.md:

## NinjaTrader Docs
Full SDK reference is in ./path/to/this/repo/. When writing NinjaScript,
consult the relevant .md files for accurate API signatures and examples.

OpenCode

Add the knowledge base path to your AGENTS.md file:

## NinjaTrader Documentation
When working with NinjaScript, reference the documentation in ./path/to/ninjatrader-kb/.
- `docs/` - API reference
- `best-practices/index.md` - Coding standards
- `education/index.md` - Tutorials

Or use the instructions field in your opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "instructions": ["./path/to/ninjatrader-kb/CLAUDE.md"]
}

Other AI Tools

Point your AI assistant to this directory. The index.json file contains a machine-readable index of all pages.

Example Files for Common Tasks

Building Indicators

• docs/indicator.md - Base class overview
• docs/onstatechange.md - Lifecycle management
• docs/onbarupdate.md - Calculation logic
• docs/addplot.md - Adding visual plots
• best-practices/index.md - Coding standards

Building Strategies

• docs/strategy.md - Base class overview
• docs/managed_approach.md - Order management
• docs/enterlong.md, docs/entershort.md - Entry methods
• docs/setstoploss.md, docs/setprofittarget.md - Risk management

File Structure

All documentation files in docs/ use a flat structure with filenames derived from the original URL: