mcp-integration

Configure MCP servers with stdio, SSE, HTTP, and WebSocket transports in Claude Code plugins.

Updated Apr 9, 2026
One-click install
npx skills add https://github.com/dobrinz123/claude-skills-share --skill mcp-integration-dobrinz123
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: mcp-integration
Source: https://github.com/dobrinz123/claude-skills-share/tree/main/plugins/plugin-dev/skills/mcp-integration
Command: npx skills add https://github.com/dobrinz123/claude-skills-share --skill mcp-integration-dobrinz123

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This skill removes the complexity of exposing external services as tools inside Claude Code plugins by standardizing Model Context Protocol (MCP) server configuration, authentication, tool discovery, and lifecycle management so developers can integrate third-party APIs, databases, and local tools reliably.

Core Features & Use Cases

  • Multi-transport support: Guidance for stdio (local processes), SSE, HTTP, and WebSocket MCP servers to support local tools, hosted services, REST APIs, and real-time streams.
  • Authentication patterns: Clear instructions for OAuth flows, bearer tokens, dynamic header helpers, and environment-variable-based credentials for secure access.
  • Lifecycle & tooling: Best practices for .mcp.json or plugin.json configuration, environment variable expansion (including ${CLAUDE_PLUGIN_ROOT}), tool naming conventions, pre-allowing tools in commands, startup/restart behavior, and error handling.
  • Testing & debugging: Local testing checklist, use of /mcp for discovery, and recommendations for debug logging, batching, and performance considerations.
  • Use case: Bundle a local database MCP server for data queries, add an Asana SSE integration for task management, or connect to a third-party REST MCP backend for automated workflows.

Quick Start

Add a .mcp.json to the plugin root with your server entry, document required environment variables, enable the plugin locally, and run /mcp to verify tool discovery and authentication.

Frequently Asked Questions about mcp-integration

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I configure an MCP server in a Claude Code plugin?

To configure an MCP server in a Claude Code plugin, add a .mcp.json file to the plugin root with your server entry, document required environment variables, and enable the plugin locally. This setup standardizes tool discovery and establishes external service connections.

What authentication methods are supported for Model Context Protocol servers?

Model Context Protocol servers support multiple authentication methods including OAuth flows, bearer tokens, dynamic header helpers, and environment-variable-based credentials. These patterns ensure secure access when integrating external APIs and hosted services into Claude Code plugins.

Can I use stdio, SSE, HTTP, and WebSocket transports with MCP servers?

Yes, MCP server integration supports stdio for local processes, SSE for hosted services, HTTP for REST APIs, and WebSocket for real-time streams. This multi-transport support allows developers to connect diverse external tools and services reliably within Claude Code plugins.

Why do I need environment variable substitution like ${CLAUDE_PLUGIN_ROOT} for MCP servers?

Environment variable substitution like ${CLAUDE_PLUGIN_ROOT} is needed for MCP servers to securely manage credentials and dynamically resolve file paths within the plugin lifecycle. It ensures portable configuration across different environments without hardcoding sensitive authentication tokens.

What is the best way to test and debug MCP tool discovery?

The best way to test and debug MCP tool discovery is to run the /mcp command locally after enabling your plugin. This verifies tool availability and authentication flows, while debug logging helps identify lifecycle issues or performance bottlenecks in external service connections.

How do I pre-allow MCP tools in Claude Code commands and agents?

To pre-allow MCP tools in Claude Code commands and agents, define the allowed tools within your plugin.json configuration. This ensures secure, synchronous, or real-time tool invocation during automated workflows without requiring manual approval during execution.