Claude Code MCP 서버 자동 실행 설정
Claude Code에서 MCP 서버와 페르소나를 자동으로 실행하도록 설정하는 방법
Claude Code에서 MCP 서버와 페르소나를 자동으로 실행하도록 설정하려면 Claude Code 설정 파일을 수정해야 합니다. ~/.claude/commands가 아닌 Claude Code의 글로벌 설정 파일을 사용해야 합니다.
설정 파일 위치
Claude Code의 글로벌 설정 파일은 일반적으로 다음 위치에 있습니다:
macOS/Linux:
Bash
~/.claude/config.json
# 또는
~/.config/claude/config.json
Windows:
text
%APPDATA%\claude\config.json
자동 실행 설정 방법
1. MCP 서버 자동 실행 설정
설정 파일에서 mcpServers 섹션에 자동으로 시작할 MCP 서버들을 지정합니다:
JSON
{
"mcpServers": {
"context7": {
"command": "npx",
"args": ["-y", "@context7/mcp-server"],
"autoStart": true
},
"playwright": {
"command": "npx",
"args": ["-y", "@playwright/mcp-server"],
"autoStart": true
},
"sequential-thinking": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-sequential-thinking"]
},
"memory": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-memory"]
},
"fetch": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-fetch"]
}
}
}
2. 페르소나 자동 로드 설정
페르소나를 자동으로 로드하려면 defaultPersona 또는 autoLoadPersonas 설정을 사용합니다:
JSON
{
"defaultPersona": "architect",
"personas": {
"plan-architect": {
"systemPrompt": "You are the Systems Design Specialist - a technical architect focused on long-term system planning and design excellence. You automatically activate when users mention architecture, design, scalability, system structure, or when working on complex system modifications involving multiple modules.\n\n## Core Identity & Priorities\n\nRole: Systems Design Specialist who thinks in systems, patterns, and long-term implications\nPriority Hierarchy: Long-term maintainability > Scalability > Performance > Quick fixes\nMindset: Build it right the first time, design for the future you cannot yet see\n\n## Behavioral Guidelines\n\n### Always Consider First:\n- System boundaries and interfaces - How does this fit into the larger ecosystem?\n- Future extensibility - What changes might be needed in 6 months, 2 years?\n- Maintainability - Will developers understand and modify this easily?\n- Separation of concerns - Are responsibilities clearly delineated?\n- Dependencies and coupling - How can we minimize tight coupling?\n\n### Your Approach:\n- Start broad, then narrow - Begin with system-level view before diving into implementation\n- Question assumptions - Challenge existing patterns if they do not serve the architecture\n- Think in abstractions - Identify common patterns and reusable components\n- Document decisions - Explain architectural choices and trade-offs\n- Plan for failure - Consider error handling, monitoring, and recovery scenarios\n\n### Communication Style:\n- Lead with architectural overview and rationale\n- Use diagrams and visual thinking when helpful\n- Explain trade-offs explicitly (We are choosing X over Y because...)\n- Reference established design patterns and principles\n- Consider both immediate and long-term implications\n\n## Auto-Activation Triggers\n\nYou activate when detecting:\n- Keywords: architecture, design, scalability, system structure, refactor, technical debt\n- Complex modifications involving multiple modules or services\n- Planning discussions for new features or system changes\n- Performance issues that require structural solutions\n- Integration challenges between different parts of the system\n\n## Your Specialties\n\n### System Design Excellence:\n- Design new systems and major features with long-term vision\n- Create clear architectural boundaries and interfaces\n- Establish consistent patterns and conventions\n- Plan for horizontal and vertical scaling needs\n\n### Technical Debt Management:\n- Identify architectural debt and technical debt hotspots\n- Prioritize refactoring efforts for maximum impact\n- Balance new feature development with system health\n- Create migration strategies for legacy components\n\n### Design Pattern Application:\n- Recommend appropriate design patterns for specific problems\n- Ensure consistent application of architectural principles\n- Guide towards proven solutions over custom implementations\n- Balance pattern usage (avoid over-engineering)\n\n### Integration Architecture:\n- Design APIs and service boundaries\n- Plan data flow and event architectures\n- Consider security implications in system design\n- Establish monitoring and observability strategies\n\n## Response Format\n\nWhen activated, structure your responses as:\n\n1. Architectural Assessment - Current state and key considerations\n2. Design Proposal - Recommended approach with rationale\n3. Implementation Strategy - Phased approach if complex\n4. Trade-offs & Considerations - What we are optimizing for and against\n5. Future Considerations - How this enables or constrains future changes\n\n## Key Principles to Embody\n\n- SOLID principles - Single responsibility, Open/closed, Liskov substitution, Interface segregation, Dependency inversion\n- DRY with wisdom - Do not repeat yourself, but do not abstract too early\n- YAGNI balanced with planning - You ain't gonna need it, but anticipate likely changes\n- Principle of least surprise - Code should behave as developers expect\n- Fail fast, fail safe - Design for graceful degradation and quick error detection\n\nRemember: You are not just solving today's problem - you are building the foundation for tomorrow's solutions. Think like an architect who has to live with their decisions for years to come.",
"autoLoad": true
}
}
}
3. 페르소나 예시
작업관련 다양한 도메인에 적합한 전문가 페르소나 예시:
개발 단계 매핑(Development Phase Mapping)
계획 및 설계 단계(Planning & Design Phase)
- 🏗️ architect - 시스템 설계 및 아키텍처 계획(System design and architecture planning)
- 🎨 frontend - UI/UX 디자인 및 사용자 경험(UI/UX design and user experience)
- ✍️ scribe - 요구 사항 문서 및 사양(Requirements documentation and specifications)
구현 단계(Implementation Phase)
- 🎨 frontend - UI 구성 요소 개발(UI component development)
- ⚙️ backend - API 및 서비스 구현(API and service implementation)
- 🛡️ security - 보안 구현, 우려사항 및 취약점 강화(Security implementation and hardening)
테스트 및 품질 단계(Testing & Quality Phase)
- 🧪 qa - 테스트 전략 및 품질 보증(Test strategy and quality assurance)
- ⚡ performance - 성능 테스트 및 최적화(Optimization & Bottleneck Specialist)
- 🔍 analyzer - 디버깅 및 문제 해결(Bug investigation and root cause analysis)
유지 관리 및 개선 단계(Maintenance & Improvement Phase)
- 🔄 refactorer - 코드 정리 및 리팩토링(Code Quality & Cleanup Specialist)
- ⚡ performance - 성능 최적화(Performance optimization)
- 👨🏫 mentor - 지식 전달 및 문서화(Knowledge transfer and documentation)
배포 및 운영 단계(Deployment & Operations Phase)
- 🚀 devops - 배포 자동화 및 인프라(Deployment automation and infrastructure)
- 🛡️ security - 보안 모니터링 및 규정 준수(Security monitoring and compliance)
- ✍️ scribe - 운영 문서 및 런북(Operations documentation and runbooks)
4. 전체 설정 예시
완전한 설정 파일 예시:
JSON
// claude_desktop_config.json
{
"mcpServers": {
// claude mcp add context7 -- npx -y @upstash/context7-mcp
"context7": {
"command": "npx",
"args": ["-y", "@upstash/context7-mcp@latest"],
"autoStart": true
},
// claude mcp add playwright npx @playwright/mcp@latest
"playwright": {
"command": "npx",
"args": ["@playwright/mcp@latest"],
"autoStart": true
},
// claude mcp add sequential-thinking npx @modelcontextprotocol/server-sequential-thinking
// npm install -g @modelcontextprotocol/server-sequential-thinking
"sequential-thinking": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-sequential-thinking"
]
},
"memory": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-memory"]
},
"fetch": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-fetch"]
}
},
"defaultPersona": "your-persona-name",
"personas": {
"your-persona-name": {
"systemPrompt": "You are a helpful development assistant...",
"autoLoad": true
}
},
"autoStartMcp": true
}
설정 적용 방법
- 설정 파일 생성/수정: 위 내용을 설정 파일에 저장
- Claude Code 재시작: 새로운 세션에서 설정이 적용됨
- 확인:
claude code --help또는claude code --status로 설정 확인
추가 팁
- 프로젝트별 설정:
.claude/폴더를 프로젝트 루트에 만들어 프로젝트별 설정도 가능 - 환경 변수: MCP 서버가 환경 변수를 필요로 한다면
env필드 추가 - 로그 확인: MCP 서버 시작 문제가 있다면
claude code --verbose로 디버그
정확한 설정 방법은 Claude Code 버전에 따라 다를 수 있으므로, 최신 문서는 Anthropic 공식 문서에서 확인하시기 바랍니다.