toc-agents

한국어 README

The complete Goldratt Theory of Constraints framework as Claude Code skills.

The first open-source implementation of TOC Thinking Processes, operations tools, and decision frameworks as AI agent skills. No external dependencies — copy the .claude/ folder into any project and start analyzing.

What is TOC?

The Theory of Constraints (TOC), created by Eliyahu M. Goldratt, is a management philosophy that views any system as limited by a small number of constraints. TOC provides systematic tools for identifying constraints, resolving conflicts, and driving continuous improvement.

"Any improvement NOT at the constraint is an illusion." — Goldratt, The Goal

Available Tools

Thinking Processes (Problem Solving)

Command	Tool	What it does	Based on
/toc "problem"	Full Workflow	CRT → EC → FRT → PRT → TT pipeline	All books
/toc:crt	Current Reality Tree	Find root causes from symptoms	It's Not Luck
/toc:ec	Evaporating Cloud	Resolve conflicts by breaking assumptions	It's Not Luck
/toc:frt	Future Reality Tree	Validate solutions, find side-effects	It's Not Luck
/toc:prt	Prerequisite Tree	Identify obstacles and intermediate objectives	It's Not Luck
/toc:tt	Transition Tree	Step-by-step action plan	It's Not Luck

Operations & Strategy

Command	Tool	What it does	Based on
/toc:five-steps	Five Focusing Steps	Identify and exploit the bottleneck	The Goal
/toc:dbr	Drum-Buffer-Rope	Pull-based scheduling around constraint	The Goal / The Race
/toc:ccpm	Critical Chain PM	Project planning with shared buffers	Critical Chain
/toc:throughput	Throughput Accounting	T/I/OE decision framework	The Haystack Syndrome
/toc:buy-in	Layers of Resistance	Overcome resistance to change (6 layers)	It's Not Luck

Installation

Option 1: Copy into your project

# Clone this repo
git clone https://github.com/ironyjk/toc-agents.git

# Copy .claude/ folder into your project
cp -r toc-agents/.claude/commands/toc* your-project/.claude/commands/
cp -r toc-agents/.claude/skills/toc your-project/.claude/skills/

Option 2: Use directly

cd toc-agents
claude   # Start Claude Code in this directory

Then invoke any command:

/toc "Our projects keep finishing late despite having enough resources"

Quick Start

Resolve a Conflict

/toc:ec "We need to cut costs to survive, but we also need to invest in quality to keep customers"

Output: Evaporating Cloud diagram with broken assumptions and injection.

Find Root Causes

/toc:crt "Delivery is late, quality is dropping, employees are quitting, customers are complaining"

Output: Current Reality Tree tracing all symptoms to 1-2 root causes.

Full Analysis Pipeline

/toc "Revenue has plateaued at $10M despite growing the team from 20 to 30 people"

Output: Complete analysis through all 5 stages — root causes, conflict resolution, solution validation, and action plan.

Evaluate a Decision

/toc:throughput "Should we accept this $50K order that uses 20 hours of our bottleneck machine?"

Output: Throughput Accounting analysis with T/I/OE comparison.

How the Full Workflow Works

graph LR
    P["Problem"] --> CRT["1. CRT<br/>Root Causes"]
    CRT --> EC["2. EC<br/>Resolve Conflict"]
    EC --> FRT["3. FRT<br/>Validate Solution"]
    FRT --> PRT["4. PRT+TT<br/>Action Plan"]
    
    style CRT fill:#d63031,color:#fff
    style EC fill:#e17055,color:#fff
    style FRT fill:#00b894,color:#fff
    style PRT fill:#0984e3,color:#fff

Loading

The full workflow uses a 4-agent sequential pipeline:

1. CRT Agent (Sonnet) — Identifies root causes from symptoms
2. EC Agent (Opus) — Resolves the core conflict by breaking hidden assumptions
3. FRT Agent (Opus) — Validates the solution and trims negative side-effects
4. PRT+TT Agent (Sonnet) — Creates an implementation roadmap with detailed steps

Each agent's output feeds into the next. The coordinator synthesizes everything into a final report.

Output Formats

All tools output in Mermaid (default) or ASCII format:

/toc:ec "speed vs quality" --format ascii

Mermaid diagrams render natively in GitHub, VS Code, and most Markdown viewers.

Theory Background

See docs/goldratt-background.md for a comprehensive overview of TOC, Goldratt's books, and the intellectual foundations.

Examples

• Business Dilemma — Classic cost vs. quality conflict (EC)
• Production Bottleneck — Manufacturing throughput (Five Steps + DBR)
• Software Architecture — Monolith vs. microservices (Full Workflow)
• Startup Scaling — Growth vs. culture (EC + CCPM)

Roadmap

v1.0

• 6 Thinking Process tools (CRT, EC, FRT, PRT, TT, Five Steps)
• 4 Operations/Strategy tools (DBR, CCPM, Throughput Accounting, Buy-in)
• Full workflow pipeline with multi-agent orchestration
• Mermaid + ASCII output

v2.0 (Current)

• OR-Tools solvers in solvers/ directory:
  • product_mix.py — Linear Programming for optimal product mix (/toc:throughput --solve)
  • dbr_scheduler.py — CP-SAT job scheduling around the drum (/toc:dbr --optimize)
  • ccpm_leveler.py — Multi-project resource leveling with staggering (/toc:ccpm --optimize)
  • buffer_sim.py — Monte Carlo buffer sizing (3 methods compared)

v3.0 (Planned)

• Strategy & Tactics Tree (S&T Tree) — Goldratt's strategic planning tool
• Viable Vision analysis — "Can every company achieve viable vision?"
• Interactive mode — step-by-step guided analysis with user input at each phase
• MCP server — expose TOC tools as MCP resources for cross-project use

Why OR-Tools?

Several TOC concepts have precise mathematical formulations:

TOC Concept	Mathematical Model	OR-Tools Solver
Product Mix (T/CU ranking)	Linear Programming	pywraplp
DBR Scheduling	Constraint Programming	cp_model (CP-SAT)
CCPM Resource Leveling	Resource-Constrained Scheduling	cp_model
Buffer Sizing	Monte Carlo Simulation	Python + numpy
Routing (multi-site)	Vehicle Routing Problem	routing

The Thinking Processes (CRT, EC, FRT, etc.) are qualitative reasoning tools best handled by LLMs. The operations tools (DBR, CCPM, Throughput Accounting) benefit from combining LLM reasoning with mathematical optimization. v2.0 will bridge both worlds.

Contributing

Contributions welcome! Areas where help is needed:

1. Domain-specific examples — manufacturing, healthcare, education, government
2. OR-Tools integration — mathematical optimization for DBR scheduling and product mix
3. Translations — Korean, Japanese, Chinese, German, Portuguese
4. Testing — run the tools against real problems and report quality

Use with /think (30 Tools)

TOC works best when combined with other frameworks. Install strategy-frameworks to get /think — a meta-agent that auto-selects the best tool(s) from 30 frameworks including TOC.

# Install all 30 tools (TOC + TRIZ + 9 strategy frameworks + /think)
curl -fsSL https://raw.githubusercontent.com/ironyjk/strategy-frameworks/master/install.sh | bash

Related repos:

• strategy-frameworks — Wardley, OODA, Porter, Blue Ocean, Design Thinking, Drucker, BSC, First Principles + /think meta-agent
• triz-agents — TRIZ (Altshuller) 9 tools for inventive problem solving

Author

Heechul Choi — CEO, DY Industrial Development
Co-created with Claude Code (Anthropic Claude Opus 4.6)

License

MIT

Acknowledgments

• Eliyahu M. Goldratt (1947–2011) — creator of the Theory of Constraints
• The TOC community for decades of practice and refinement
• Inspired by autoresearch — autonomous iteration framework for Claude Code