mcp-gateway

The mcp-gateway skill provides an operational framework for engineers working with the Model Context Protocol (MCP). It is designed for developers who need to run GitHub MCP servers locally or within containerized environments, ensuring seamless integration between development tools like the GitHub Copilot CLI and authenticated proxy gateways. This skill simplifies the lifecycle management of MCP containers, allowing users to start, verify, debug, and terminate gateway instances with precision.

• Automated deployment of MCP Gateway (gh-aw-mcpg) containers via python scripts.

• Support for multiple transport modes including HTTP, SSE, and stdio-based communication.

• Configuration management for GitHub Personal Access Tokens (PATs) and environment-specific variables like GATEWAY_PORT and GATEWAY_API_KEY.

• Health monitoring and log analysis for detecting token passthrough failures or container exit errors.

• Seamless integration with the Agentic Workflow Firewall (AWF) to ensure secure host-access and domain filtering.

• Native support for testing MCP tool registration and troubleshooting connectivity errors like ECONNREFUSED.

• Always pipe valid JSON configurations via stdin when spawning gateway containers to maintain compatibility with the v0.0.59+ schema.

• Use the provided scripts such as start-gateway.py and debug-gateway.py for consistent operational status reports.

• Ensure the Docker socket is correctly mounted for lifecycle control of nested MCP server containers.

• Always set GITHUB_PERSONAL_ACCESS_TOKEN explicitly in your configuration to avoid authentication scope issues.

• When using with AWF, remember to enable --enable-host-access and mount the necessary volumes for the configuration files.