dcisionai-mcp-server 3.0.0

DcisionAI MCP Server v3.0 - AI-Powered Optimization with DAME (DcisionAI Micro-differential Evolutionary Algorithm), HiGHS Solver, Dual Validation, 90%+ Trust Scores & Business Interpretation

🎯 DcisionAI MCP Server

AI-Powered Optimization for Your IDE - Solve complex business problems directly in Cursor, Claude Desktop, or VS Code.

---

✨ What is This?

DcisionAI is a Model Context Protocol (MCP) server that brings enterprise-grade optimization to your IDE. Ask your AI assistant to solve real business problems, and get:

• 🎯 90%+ Trust Scores with mathematical proof
• 🧠 DAME Algorithm (proprietary evolutionary solver)
• ⚖️ Dual Validation with HiGHS exact solver
• 💼 Business Interpretation in plain language
• 📊 5-Layer Proof Suite for transparency

---

🚀 Quick Start (30 seconds)

Step 1: Install in Your IDE

For Cursor / Claude Desktop:

Add to your MCP settings (~/Library/Application Support/Claude/claude_desktop_config.json on Mac):

{
  "mcpServers": {
    "dcisionai": {
      "command": "uvx",
      "args": ["dcisionai-mcp-server@latest"],
      "env": {
        "OPENAI_API_KEY": "${OPENAI_API_KEY}",
        "ANTHROPIC_API_KEY": "${ANTHROPIC_API_KEY}",
        "SUPABASE_URL": "https://your-project.supabase.co",
        "SUPABASE_KEY": "your-anon-key"
      },
      "autoApprove": ["dcisionai_solve"]
    }
  }
}

That's it! No compilation, no setup, no auth tokens.

---

Step 2: Try It

Ask Claude/Cursor:

"Use DcisionAI to optimize my retail store layout: 
20 products across 5 shelves, maximize revenue and customer flow."

Get back:

• ✅ Optimal product placement
• ✅ 90% trust score with mathematical proof
• ✅ Business interpretation
• ✅ Constraint verification
• ✅ Sensitivity analysis

---

🎓 What Can It Solve?

📊 Finance

• Portfolio Optimization - Rebalance $500k portfolio, reduce concentration risk
• Trading Schedule - Optimize execution timing to minimize market impact
• Asset Allocation - Balance risk/return across sectors

🏪 Retail

• Store Layout - Optimize shelf space for 20+ products
• Promotion Scheduling - Maximize revenue with budget constraints
• Inventory Placement - Minimize stockouts & overstock

🚚 Logistics

• Vehicle Routing - Minimize travel time for 10+ delivery stops
• Warehouse Layout - Optimize picking paths
• Job Shop Scheduling - Sequence 15 jobs across 5 machines

👷 Workforce

• Shift Scheduling - Assign 20 workers to 40 shifts
• Maintenance Scheduling - Minimize downtime across 10 assets
• Skill Matching - Optimal worker-task assignment

---

🔧 Environment Variables

Required in your IDE's MCP config:

# LLM APIs (required)
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...

# Supabase (required - stores domain configs)
SUPABASE_URL=https://your-project.supabase.co
SUPABASE_KEY=eyJ...

# Optional: External data (for real-world augmentation)
POLYGON_API_KEY=...              # Market data
ALPHA_VANTAGE_API_KEY=...        # Economic data

---

📊 Trust & Validation

Every solution includes:

1. Mathematical Proof Suite (5 proofs)

• Constraint Verification - All business rules satisfied?
• Monte Carlo Simulation - 1000 scenarios, how stable?
• Optimality Certificate - How close to theoretical best?
• Sensitivity Analysis - What if inputs change ±20%?
• Benchmark Comparison - Beats naive baseline by X%?

2. Dual Solver Validation

• DAME (heuristic) - Fast, handles any problem
• HiGHS (exact) - Slow, LP/MIP only, mathematically optimal
• Cross-validation - Compare results, boost trust to 95%+

3. Business Interpretation

• Plain language explanation
• Implementation steps
• Risks & assumptions
• What-if scenarios

---

💡 Example Interaction

You ask:

"I have 50 products to place on 8 store shelves. 
Products have different profit margins and sales rates. 
Dairy needs refrigeration. High-value items need security.
How should I arrange them?"

DcisionAI returns:

✅ Status: SUCCESS
📊 Industry: RETAIL
🎯 Domain: Store Layout Optimization
⭐ Trust Score: 92% (VERIFIED)
🏆 Certification: VERIFIED

📈 Objective Value: 0.423 (42.3% improvement vs. baseline)

💼 Business Interpretation:
"Strategic product placement optimizes for revenue and customer 
flow. High-margin products positioned in prime visibility zones. 
Refrigerated items grouped for efficiency..."

🔍 Mathematical Proof:
✅ All 8 constraints satisfied
✅ Monte Carlo: 1000 scenarios, 95% confidence
✅ Optimality: Within 4.2% of theoretical best
✅ Sensitivity: Stable under ±20% demand changes
✅ Benchmark: 42% better than random placement

🛠️ Implementation Steps:
1. Reorganize shelf 1-3 (high-traffic zone)
2. Group dairy in refrigerated section
3. Position security items near checkout
4. Monitor sales for 30 days and adjust

---

🏗️ Architecture

User Question
    ↓
LLM Extraction (GPT-4/Claude)
    ↓
Domain Classification (11 domains)
    ↓
Data Augmentation (synthetic + external APIs)
    ↓
Parallel Solving
    ├─ DAME (heuristic, 100 generations)
    └─ HiGHS (exact, LP/MIP)
    ↓
Cross-Validation
    ↓
Proof Generation (5 proofs)
    ↓
Business Interpretation (LLM)
    ↓
Structured Response (JSON)

---

📚 Supported Domains

Domain	Description	Example
Portfolio	Asset allocation, risk balancing	"Optimize $500k portfolio"
Retail Layout	Shelf space allocation	"Place 20 products on 5 shelves"
VRP	Vehicle routing, delivery optimization	"Route 3 trucks to 15 stops"
Job Shop	Production scheduling	"Schedule 10 jobs on 4 machines"
Workforce	Shift assignment, rostering	"Assign 15 workers to 40 shifts"
Maintenance	Asset maintenance scheduling	"Schedule maintenance for 8 machines"
Promotion	Marketing campaign optimization	"Allocate $50k ad budget"
Trading	Trade execution optimization	"Minimize market impact of large order"
Customer Onboarding	Wealth management	"Onboard new client with $2M"
PE Exit Timing	Private equity exits	"Optimize exit timing for 5 holdings"
HF Rebalancing	Hedge fund portfolio adjustments	"Rebalance multi-strategy fund"

---

🔬 Technology Stack

• DAME - Proprietary evolutionary algorithm (see research paper)
• HiGHS - Open-source LP/MIP solver from Edinburgh
• FastMCP - Model Context Protocol implementation
• Anthropic Claude - LLM for extraction & interpretation
• OpenAI GPT-4 - Fallback LLM
• Supabase - Domain config storage

---

📖 Documentation

• Research Paper: DcisionAI.pdf
• GitHub: dcisionai/dcisionai-mcp-platform
• Issues: Report a bug

---

🛠️ Development

Local Testing

# Clone the repo
git clone https://github.com/dcisionai/dcisionai-mcp-platform.git
cd dcisionai-mcp-platform/dcisionai/fastapi-server

# Install dependencies
pip install -e .

# Run locally
python -m dcisionai_mcp_server.server

Contributing

We welcome contributions! See CONTRIBUTING.md.

---

📜 License

MIT License - see LICENSE

---

🤝 Support

• Email: [e-mail]
• GitHub Issues: Report a bug
• Docs: Full documentation

---

🎉 Credits

Built by Amey Dhavle

Powered by:

• FastMCP
• HiGHS Solver
• Anthropic Claude
• OpenAI GPT-4
• Supabase

---

⭐ Star us on GitHub if DcisionAI helps your business!

github.com/dcisionai/dcisionai-mcp-platform