MCP Integration FAQs

Question 1

What types of MCP server protocols are supported?

Accepted Answer

This skill supports four major transport protocols: stdio (for local child processes), SSE (Server-Sent Events for cloud services with OAuth), HTTP (for RESTful APIs with token auth), and WebSocket (for real-time bidirectional communication).

Question 2

How are MCP tools named within Claude Code?

Accepted Answer

Tools provided via MCP are automatically namespaced to avoid conflicts. They follow the format: 'mcp__plugin____'. You can then pre-allow these specific tools in your command frontmatter for secure, authorized access.

Question 3

How do I configure an MCP server in my plugin?

Accepted Answer

You can configure MCP servers using two primary methods: creating a dedicated '.mcp.json' file at your plugin root for better organization, or adding an 'mcpServers' field directly into your 'plugin.json'. Both methods support environment variable expansion for secure and portable credential management.

Question 4

What is the MCP Integration skill for Claude Code?

Accepted Answer

The MCP Integration skill allows Claude Code plugins to connect with external services and APIs using the Model Context Protocol (MCP). It provides a standardized way to expose external tools, databases, and file systems directly to Claude, enabling the AI to perform actions across your entire tech stack.

Question 5

How does PulseMCP help with this skill?

Accepted Answer

PulseMCP is integrated into this skill to facilitate the discovery of over 6,800 existing MCP servers. It allows you to search for official and community-built servers for services like GitHub, Asana, and Slack, providing ready-to-use configuration patterns for your plugins.

MCP Integration

MCP Integration

Key Features

Use Cases

Key Features

Use Cases