Gurddy

You run an MCP server crafted to solve CSP, LP, minimax, and SciPy-powered optimization problems. It provides a unified protocol transport that can be used locally in IDEs or over HTTP for web clients, making it easy to solve a wide range of constrained problems with consistent tooling.

How to use

You connect a client that speaks the MCP protocol to the Gurddy MCP Server. Use the stdio transport for local IDE integration or the HTTP transport for web-based access. With the same set of tools, you can explore classic CSP problems (like N-Queens and Sudoku), LP/MIP optimization, game theory scenarios, and SciPy-powered optimizations. Start by choosing a transport, then load the available tools, and invoke the ones that match your problem and constraints.

In practice, you will set up a client that can send JSON-RPC 2.0 requests to either the stdio interface or the HTTP endpoint. You will use the provided tool names (info, install, run_example, solve_n_queens, solve_sudoku, solve_lp, solve_minimax_game, and more) to query capabilities, install or upgrade the Gurddy package, run example problems, and perform your own optimizations.

How to install

Prerequisites you need before installation: Python and a suitable package manager. You also have options for running locally via pip or using a dynamic runtime like uvx for always-up-to-date execution, or you can run a HTTP server container for remote access.

Install the Gurddy MCP package from PyPI (recommended for a straightforward setup):

pip install gurddy_mcp

# Optional development dependencies
pip install gurddy_mcp[dev]

If you want to run the MCP server via the uvx workflow (recommended to always get the latest published version):

# Prerequisites for uvx
# macOS/Linux
curl -LsSf https://astral.sh/uv/install.sh | sh

# Or using pip
pip install uv

# Or using Homebrew (macOS)
brew install uv

Then configure in your IDE with the uvx-based command to load the Gurddy MCP server.

If you already have gurddy-mcp installed via Python, you can run the local MCP stdio server directly with its command after installation:

echo '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}' | gurddy-mcp

## Additional sections

Configuration and deployment notes help you tailor the server to your environment. For local IDE usage, you typically integrate via the stdio channel and load the Gurddy MCP server as a tool in your IDE configuration. For remote access or web clients, run the HTTP transport (MCP over HTTP) and point your client at the HTTP endpoint. The HTTP server exposes a root endpoint for general access, a health check, and a dedicated /mcp/http endpoint for JSON-RPC 2.0 requests with optional streaming.

Security and trust: you can auto-approve trusted tools to streamline execution during development. When you deploy in production, review the default auto-approval and scope it to known tools you control.

Examples show real-world problem solving across CSP, LP, minimax, and SciPy contexts. Run example problems to understand limits and performance you can expect, then adapt the tool invocations to your own data sets and constraints.

Troubleshooting common issues: ensure dependencies are installed, verify the correct server transport is running, and confirm the client is issuing valid MCP tool calls. If a specific error appears, review the related tool's requirements and inputs, such as the dimensionality and type of your data.

## Notes on deployment and tooling

Docker deployment is supported. Build the image and run the container to expose the HTTP endpoint on port 8080, then access the root or the /mcp/http path from your client.

The server provides a range of problems and tools, from classic math puzzles to CSP instances and SciPy-powered optimization. Use the tool list to discover available endpoints, and call the tools to solve the problems you are interested in.

## Available tools

### info

Get information about Gurddy MCP server capabilities and status

### install

Install or upgrade the Gurddy package and runtime dependencies

### run\_example

Run predefined example problems to illustrate capabilities (n\_queens, graph\_coloring, minimax, scipy\_optimization, classic\_problems)

### solve\_n\_queens

Solve the N-Queens problem for a given board size N

### solve\_sudoku

Solve a standard 9×9 Sudoku puzzle using CSP

### solve\_graph\_coloring

Solve graph coloring problems with a configurable number of colors

### solve\_map\_coloring

Solve region coloring problems with adjacency constraints

### solve\_lp

Solve Linear Programming and Mixed Integer Programming problems

### solve\_production\_planning

Solve production planning problems with optional sensitivity analysis

### solve\_minimax\_game

Solve two-player zero-sum games and find equilibria

### solve\_minimax\_decision

Robust optimization to minimize maximum loss under scenarios

### solve\_24\_point\_game

Solve the 24-point arithmetic game for given numbers

### solve\_chicken\_rabbit\_problem

Solve the classic chicken-rabbit head-and-legs problem

### solve\_scipy\_portfolio\_optimization

Nonlinear portfolio optimization using SciPy

### solve\_scipy\_statistical\_fitting

Statistical parameter estimation using SciPy with distribution fitting

### solve\_scipy\_facility\_location

Facility location optimization using a hybrid CSP-SciPy approach