OR Tools

🚀 MCP-ORTools

An MCP (Model Context Protocol) server implemented using Google OR-Tools constraint programming solver. Specifically designed for large language models by standardizing constraint model specifications.

🚀 Quick Start

MCP-ORTools integrates Google's OR-Tools constraint programming solver with large language models through the Model Context Protocol, enabling AI models to:

• Submit and validate constraint models
• Set model parameters
• Solve constraint satisfaction and optimization problems
• Obtain and analyze solutions

📦 Installation

1. Install the package:

pip install git+https://github.com/Jacck/mcp-ortools.git

2. Configure Claude Desktop Create a configuration file at %APPDATA%\Claude\claude_desktop_config.json (Windows) or ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):

{
  "mcpServers": {
    "ortools": {
      "command": "python",
      "args": ["-m", "mcp_ortools.server"]
    }
  }
}

📚 Documentation

Model Specification

The model is specified in JSON format and consists of three main parts:

• variables: Define variables and their domains.
• constraints: A list of constraints using OR-Tools method syntax.
• objective: An optional optimization objective.

Constraint Syntax

Constraints must use OR-Tools method syntax:

• .__le__() represents less than or equal to (<=).
• .__ge__() represents greater than or equal to (>=).
• .__eq__() represents equal to (==).
• .__ne__() represents not equal to (!=).

💻 Usage Examples

Basic Usage

Simple Optimization Model

{
    "variables": [
        {"name": "x", "domain": [0, 10]},
        {"name": "y", "domain": [0, 10]}
    ],
    "constraints": [
        "(x + y).__le__(15)",
        "x.__ge__(2 * y)"
    ],
    "objective": {
        "expression": "40 * x + 100 * y",
        "maximize": true
    }
}

Advanced Usage

Knapsack Problem

Example: Select items with values [3,1,2,1] and weights [2,2,1,1], with a total weight limit of 2.

{
    "variables": [
        {"name": "p0", "domain": [0, 1]},
        {"name": "p1", "domain": [0, 1]},
        {"name": "p2", "domain": [0, 1]},
        {"name": "p3", "domain": [0, 1]}
    ],
    "constraints": [
        "(2*p0 + 2*p1 + p2 + p3).__le__(2)"
    ],
    "objective": {
        "expression": "3*p0 + p1 + 2*p2 + p3",
        "maximize": true
    }
}

Additional Constraint Example

{
    "constraints": [
        "p0.__eq__(1)",         // Item p0 must be selected
        "p1.__ne__(p2)",        // p1 and p2 cannot be selected simultaneously
        "(p2 + p3).__ge__(1)"   // At least one of p2 or p3 must be selected
    ]
}

✨ Features

• Fully supports the OR-Tools CP-SAT solver.
• JSON-based model specification.
• Supports:
  • Integer and boolean variables (domain: [min, max]).
  • Linear constraints using OR-Tools method syntax.
  • Linear optimization objectives.
  • Timeouts and solver parameters.
  • Binary constraints and relationships.
  • Portfolio selection problems.
  • Knapsack problems.

Supported Operations in Constraints

• Basic arithmetic: +, -, *
• Comparison: .le(), .ge(), .eq(), .ne()
• Linear combinations of variables
• Binary logic through constraint combinations

🚧 Development

For development setup:

git clone https://github.com/Jacck/mcp-ortools.git
cd mcp-ortools
pip install -e .

📚 Documentation

Model Response Format

The solver will return the following JSON structure:

{
    "status": "success",           // Status: "success" or "error"
    "result": {
        "x": 5,                   // Value of variable x
        "y": 10                   // Value of variable y
    },
    "message": ""                 // Error message (if any)
}

Status Codes

• success: Operation succeeded.
• error: An error occurred.

📄 License

Apache License 2.0 - Please refer to the LICENSE file for detailed information.