mcp-optimizer

MCP-Optimizer is a Python-based MCP server that provides mathematical optimization capabilities using PuLP and OR-Tools. It supports two transport modes: STDIO for direct MCP client integration (Claude Desktop, Cursor, etc.) and SSE for remote web clients. The server can be run via uvx, Docker, or a Python/Pip installation, giving flexibility across platforms. Once running, you can connect through an MCP-compatible client and send standard MCP messages to initialize, query capabilities, and perform optimization tasks. The SSE endpoint is typically exposed at /sse when using the SSE transport.

Prerequisites:

• Python 3.11+ (for native Python usage) or a working uvx environment
• Docker (optional, for containerized runs)
• Optional: a Python package manager (pip) or uvx tooling

Option A — Docker (recommended for stability):

# Run the MCP-Optimizer container with STDIO transport by default
docker run --rm -i ghcr.io/dmitryanchikov/mcp-optimizer:latest

# Run with SSE transport exposed on port 8000
docker run -d -p 8000:8000 -e TRANSPORT_MODE=sse ghcr.io/dmitryanchikov/mcp-optimizer:latest

# Check SSE endpoint
curl -i http://localhost:8000/sse

Option B — Python with Pip (standard installation):

# Create and activate a virtual environment (optional but recommended)
python -m venv .venv
source .venv/bin/activate  # Linux/macOS
# or .venv\Scripts\activate  # Windows

# Install the package
pip install mcp-optimizer

# Run in STDIO mode (default) or specify transport
mcp-optimizer --transport stdio

Option C — uvx (universal runner):

# Install uvx if not already installed
# Then run the MCP-Optimizer via uvx
uvx mcp-optimizer

# Run with SSE transport explicitly
uvx mcp-optimizer --transport sse

Additional notes:

• If you intend to use OR-Tools on macOS, you may need to install OR-Tools via Homebrew and use pip install "mcp-optimizer[stable]" for full support.
• You can tailor the transport through environment variables or CLI options as shown above.

Tips and common issues:

• The SSE endpoint is typically http://localhost:8000/sse when using the SSE transport.
• For macOS users, the uvx pathway may have limited functionality; prefer Docker or pip installation for full PuLP and OR-Tools support.
• If you encounter transport or dependency issues, ensure your environment variable TRANSPORT_MODE is set to a supported value (stdio or sse) and that the chosen method has the appropriate dependencies installed.
• When using Claude Desktop or Cursor integrations, you can generate configuration blocks that point to the appropriate command and transport, as demonstrated in the README examples.