MCP Integration for Plugins Skill

A Claude Code skill for integrating MCP (Model Context Protocol) servers into Claude Code plugins — covering server discovery, tool registration, resource access, and multi-server orchestration.

When to Use This Skill

Choose this skill when:

• Connecting Claude Code to external MCP servers
• Configuring multiple MCP tool providers
• Building plugins that consume MCP resources
• Setting up MCP server discovery and connection
• Orchestrating tools across multiple MCP servers
• Debugging MCP connection and communication issues

Consider alternatives when:

• You need to build an MCP server (use the MCP builder skill)
• You need a REST API integration (use HTTP directly)
• You need to modify Claude Code internals (not possible)

Quick Start

Copy
// .claude/settings.json - Configure MCP servers
{
  "mcpServers": {
    "database": {
      "command": "node",
      "args": ["./mcp-servers/database-server.js"],
      "env": { "DATABASE_URL": "postgresql://localhost:5432/mydb" }
    },
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": { "GITHUB_TOKEN": "ghp_..." }
    },
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/allowed/dir"]
    }
  }
}

Copy
# Add an MCP server to Claude Code
claude mcp add database -- node ./mcp-servers/database-server.js

# List configured servers
claude mcp list

# Test a server connection
claude mcp test database

Core Concepts

MCP Architecture

Component	Role	Location
MCP Client	Claude Code's built-in connector	Inside Claude Code
MCP Server	Provides tools, resources, prompts	External process
Transport	Communication channel	stdio, SSE, HTTP
Tool	Callable function with schema	Defined in server
Resource	Read-only data endpoint	Defined in server

Server Configuration Patterns

Copy
// Local server (stdio transport)
{
  "my-server": {
    "command": "node",
    "args": ["./server.js"],
    "env": { "API_KEY": "..." }
  }
}

// Remote server (SSE transport)
{
  "remote-server": {
    "url": "https://mcp.example.com/sse",
    "headers": { "Authorization": "Bearer token" }
  }
}

// Python server
{
  "python-server": {
    "command": "python",
    "args": ["-m", "my_mcp_server"],
    "env": { "CONFIG_PATH": "./config.json" }
  }
}

Multi-Server Orchestration

Copy
// Configure multiple specialized servers
{
  "mcpServers": {
    "database": {
      "command": "node",
      "args": ["./servers/db-server.js"],
      "env": { "DB_URL": "..." }
    },
    "monitoring": {
      "command": "node",
      "args": ["./servers/monitoring-server.js"],
      "env": { "DATADOG_API_KEY": "..." }
    },
    "docs": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "./docs"]
    }
  }
}

Configuration

Parameter	Type	Default	Description
command	string	—	Command to start the MCP server
args	array	[]	Arguments passed to the command
env	object	{}	Environment variables for the server process
url	string	—	URL for remote SSE/HTTP servers
headers	object	{}	HTTP headers for remote servers
timeout	number	30000	Connection timeout in milliseconds
restart_on_failure	boolean	true	Auto-restart crashed servers

Best Practices

1. Use environment variables for secrets, not inline config — store API keys and tokens in environment variables referenced by the MCP config; never commit secrets in .claude/settings.json.

2. Keep servers focused on a single domain — a database server provides database tools, a monitoring server provides monitoring tools; don't bundle unrelated tools in one server.

3. Test servers independently before integrating — run claude mcp test <server> to verify each server works before relying on it in complex workflows.

4. Use stdio transport for local servers — stdio is the simplest and most reliable transport for servers running on the same machine; use SSE or HTTP only for remote servers.

5. Configure restart-on-failure for critical servers — MCP servers can crash due to bugs or resource exhaustion; auto-restart ensures tools remain available during long sessions.

Common Issues

Server fails to start — Check that the command is installed and accessible in your PATH. Run the command manually to see error output. Verify environment variables are set correctly.

Tools from different servers have name conflicts — Each MCP server should use unique tool names. Prefix tools with the server's domain: db_query, monitor_metrics, docs_search.

Remote server connection drops — SSE connections can timeout due to proxies or firewalls. Implement heartbeat/keepalive in your server and configure proxy timeouts appropriately.