Skyforge

You run a Model Context Protocol (MCP) server that dynamically exposes SkySpark Axon functions as MCP tools, enabling AI assistants to discover, prompt, and invoke SkySpark capabilities in a secure and scalable way. This server supports both local development (stdio) and web clients (HTTP/SSE), with automatic tool discovery from SkySpark on demand and type-safe JSON schemas for tool parameters and results.

How to use

Connect to the MCP server from your MCP client to discover available tools. The server fetches tool definitions from SkySpark when you request the tool list, so you can add new tools simply by creating Axon functions in SkySpark without restarting the MCP server.

How to install

Prerequisites you need before running the MCP server.

Choose an installation path and follow the steps below.

# Prerequisites
- Docker (recommended) OR Python 3.12+ with uvx
- SkySpark or Haxall server with API access

# Docker setup (easiest)
# 1. Clone the project and configure environment
git clone https://github.com/yourusername/skyforge-mcp.git
cd skyforge-mcp

# 2. Create environment file
cat > .env << EOF
SKYSPARK_URI=http://host.docker.internal:8080/api/demo
SKYSPARK_USERNAME=your_username
SKYSPARK_PASSWORD=your_password
EOF

# 3. Start the server
docker-compose up --build

# The server will be reachable at
# http://localhost:8000/mcp

# Local development setup (stdio or HTTP/SSE)
# 1. Install uv if you prefer a local runtime
curl -LsSf https://astral.sh/uv/install.sh | sh

# 2. Clone and set up
git clone https://github.com/yourusername/skyforge-mcp.git
cd skyforge-mcp
uv sync

# 3. Create the same .env file as above for SkySpark credentials

# 4. Run stdio mode for Claude Desktop clients
uv run main.py

# 5. Or run HTTP/SSE mode for web clients
uv run uvicorn main:app --host 0.0.0.0 --port 8000

Additional sections

Configuration details, security considerations, troubleshooting steps, and example tool integration follow practical, concrete steps so you can operate the MCP server in development and test environments with confidence.

If you plan to test locally with SkySpark, ensure your SkySpark API endpoint is reachable from your deployment and that the credentials used in your .env file are correct. For Docker, use host.docker.internal to reach the host SkySpark instance when running inside a container.

Security and troubleshooting notes

This alpha release has no built-in authentication and wide-open CORS during development. Do not expose the server to untrusted networks in production. Use a VPN, firewall, or proper reverse proxy with authentication for production deployments.

Common issues include connection errors to SkySpark, missing tools, or tool schema changes. If tools don’t appear, verify the fetchMcpTools function exists in SkySpark and check server logs with your chosen start command. For Docker, inspect logs with docker-compose logs and restart with docker-compose up --build.

Notes about tools and tool discovery

Tools are discovered dynamically from SkySpark on each list_tools request. You can add new tools by creating Axon functions in SkySpark; there is no server restart required for schema changes.

Available tools

getSiteEquips

Get all equipment for a site from SkySpark