skillsmith

skillsmith is a Python-based scaffold that can optionally expose an MCP server to retrieve and manage its modular skills at runtime. The MCP integration wires a server endpoint for on-demand access to skills and workflows (e.g., list_skills, get_skill(name), search_skills(query), and compose_workflow(goal)). To start the MCP server locally, run the CLI via Python and the serve command: python -m skillsmith serve. You can also run the server in HTTP mode by supplying transport, host, and port options, for example: python -m skillsmith serve --transport http --host localhost --port 47731. The MCP server will then expose its endpoints over the chosen transport, allowing compatible clients (like code assistants or editors) to discover and invoke skills from the skillsmith workspace.

Prerequisites:

• Python 3.8+ and pip
• Optional: virtual environment support (recommended)

Install the base package:

pip install skillsmith

Optional MCP support (enables MCP features when available):

pip install skillsmith[mcp]

Usage to start the server locally:

# Run the MCP server in stdio mode (default)
python -m skillsmith serve

# Run the MCP server over HTTP (example)
python -m skillsmith serve --transport http --host localhost --port 47731

If you prefer to run from a source checkout, ensure PYTHONPATH is set to include the package source and start with the same command:

PYTHONPATH=src python -m skillsmith serve

Tips and caveats:

• The MCP server is optional and depends on the skillsmith package; use pip install skillsmith[mcp] to enable MoC (MCP) features.
• If the skillsmith command is not found, you can still run modules directly with python -m skillsmith (as shown above).
• When using HTTP transport, ensure the host and port are accessible to your clients and consider firewall settings.
• You can customize the workspace and skills discovery via the skillsmith CLI (e.g., init, list, add, update, lint, snapshot, watch, etc.), which in turn influences what the MCP server exposes.
• If you encounter PATH issues, the doctor tool (python -m skillsmith doctor) can suggest automatic fixes.