MCP Integration

Configure MCP servers in Claude Code plugins via .mcp.json or plugin.json.

Updated Aug 27, 2026
One-click install
npx skills add https://github.com/stellarone/evaluate --skill mcp-integration-stellarone
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: MCP Integration
Source: https://github.com/stellarone/evaluate/tree/main/.agents/skills/mcp-integration
Command: npx skills add https://github.com/stellarone/evaluate --skill mcp-integration-stellarone

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Integrating external services into Claude Code can be blocked by unclear setup steps, mismatched transport/auth expectations, and tool-naming constraints that prevent safe command/agent usage.

Core Features & Use Cases

  • Connect and configure MCP servers using .mcp.json or plugin.json so tools are discovered and usable.
  • Support multiple transports (stdio, SSE, HTTP, WebSocket) and map each to the right use case such as local tooling, hosted OAuth services, REST APIs, or real-time streaming.
  • Enable secure authentication patterns including OAuth for SSE/HTTP and token/env-based auth for HTTP/stdio, with guidance for required environment variable documentation.
  • Control tool access safely by using allowed-tools to restrict exactly which MCP tools commands can call.

Use this when you need to add or configure an MCP server (e.g., “add MCP server”, “integrate MCP”, “use .mcp.json”, “set up Model Context Protocol”, or “connect external service”) so that Claude Code plugins can expose external capabilities as structured tools.

Quick Start

Add a .mcp.json file at your plugin root that defines your MCP server transport (stdio/SSE/HTTP/ws) and then verify the server and tool names by running the /mcp command in Claude Code.

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 Claude Code?

To configure an MCP server in Claude Code, define your server transport and authentication details in a `.mcp.json` or `plugin.json` file at your plugin root. You can then verify the server and tool names by running the `/mcp` command.

What transports are supported for Model Context Protocol integration?

Model Context Protocol integration supports stdio for local servers, SSE and HTTP for hosted or REST APIs, and WebSocket for real-time bidirectional streaming. Each transport maps to specific use cases like local tooling or hosted OAuth services.

Does MCP integration support OAuth and token authentication?

Yes, MCP integration enables secure authentication patterns including OAuth for SSE and HTTP transports, alongside token or environment-variable based auth for HTTP and stdio connections. Required environment variables must be documented for setup.

How can I restrict which MCP tools a command is allowed to call?

You can control tool access safely by using the `allowed-tools` configuration to restrict exactly which MCP tools your commands and agents can call. This prevents unauthorized tool execution while following MCP tool naming conventions.

When should I use stdio versus SSE for an MCP server connection?

Use stdio for connecting local servers and tooling, while SSE or HTTP is suited for hosted OAuth services and REST APIs. WebSocket transport should be applied when you need real-time bidirectional integration with external services.