Publish package to PyPI for easier installation (uvx / pip support)

[filhocf]

Problem

The huaweicloud-mcp-server package is not published on PyPI, which means users cannot install it via standard Python package managers:

# These don't work currently:
pip install huaweicloud-mcp-server
uvx huaweicloud-mcp-server

This makes it harder to integrate with MCP-compatible IDEs and tools (like Cursor, Kiro, Cline, etc.) that typically configure MCP servers using uvx for Python packages.

Currently, users must:

1. Clone the entire repository
2. Run uv run from the project root directory
3. Or install locally with pip install -e .

Proposed Solution

Publish the package to PyPI so users can:

# Install globally
pip install huaweicloud-mcp-server

# Run directly with uvx (no install needed)
uvx huaweicloud-mcp-server mcp-server-ecs -t stdio

# Or use in MCP config (e.g., for Kiro/Cursor/Cline)
{
  "mcpServers": {
    "huawei-ecs": {
      "command": "uvx",
      "args": ["huaweicloud-mcp-server", "mcp-server-ecs", "-t", "stdio"],
      "env": {
        "HUAWEI_ACCESS_KEY": "...",
        "HUAWEI_SECRET_KEY": "..."
      }
    }
  }
}

Steps Needed

1. Verify pyproject.toml - The project already has a well-structured pyproject.toml with entry points defined in [project.scripts]. This should work as-is for PyPI.

2. Set up PyPI publishing - Add a GitHub Actions workflow for automated publishing:

   # .github/workflows/publish.yml
   name: Publish to PyPI
   on:
     release:
       types: [published]
   jobs:
     publish:
       runs-on: ubuntu-latest
       steps:
         - uses: actions/checkout@v4
         - uses: actions/setup-python@v5
           with:
             python-version: '3.10'
         - run: pip install build twine
         - run: python -m build
         - run: twine upload dist/*
           env:
             TWINE_USERNAME: __token__
             TWINE_PASSWORD: ${{ secrets.PYPI_TOKEN }}

3. Test the build locally:

   pip install build
   python -m build
   pip install dist/huaweicloud_mcp_server-0.3.0-py3-none-any.whl
   mcp-server-ecs -t stdio  # Should work

4. Consider package naming - huaweicloud-mcp-server is a good name. Verify it's available on PyPI.

Benefits

• Easier adoption by the MCP community
• Standard installation workflow (pip install / uvx)
• Seamless integration with MCP-compatible IDEs
• Version management through PyPI releases
• No need to clone the full repository

Additional Context

• The pyproject.toml already defines 100+ entry points (mcp-server-ecs, mcp-server-cce, etc.)
• Dependencies are already properly declared
• The package structure with setuptools is ready for building