mcp-server-mas-sequential-thinking MCP — 사용 사례, 설치 & 라이브 데모

왜 쓰나요

핵심 기능

Six-perspective analysis: factual, emotional, critical, optimistic, creative, synthesis
Exposed as a single MCP tool: sequentialthinking
Works with DeepSeek (default), Groq, OpenRouter, GitHub Models, Anthropic, or local Ollama
Optional Exa web search integration for 4 of the agents
Dual-model strategy — stronger model for synthesis

라이브 데모

실제 사용 모습

server-mas-sequential-thinking.replay ▶ 준비됨

0/0

설치

클라이언트 선택

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "server-mas-sequential-thinking": {
      "command": "uvx",
      "args": [
        "mcp-server-mas-sequential-thinking"
      ],
      "_inferred": true
    }
  }
}

Claude Desktop → Settings → Developer → Edit Config 열기. 저장 후 앱 재시작.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "server-mas-sequential-thinking": {
      "command": "uvx",
      "args": [
        "mcp-server-mas-sequential-thinking"
      ],
      "_inferred": true
    }
  }
}

Cursor는 Claude Desktop과 동일한 mcpServers 스키마 사용. 프로젝트 설정이 전역보다 우선.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "server-mas-sequential-thinking": {
      "command": "uvx",
      "args": [
        "mcp-server-mas-sequential-thinking"
      ],
      "_inferred": true
    }
  }
}

Cline 사이드바의 MCP Servers 아이콘 클릭 후 "Edit Configuration" 선택.

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "server-mas-sequential-thinking": {
      "command": "uvx",
      "args": [
        "mcp-server-mas-sequential-thinking"
      ],
      "_inferred": true
    }
  }
}

Claude Desktop과 같은 형식. Windsurf 재시작 후 적용.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "server-mas-sequential-thinking",
      "command": "uvx",
      "args": [
        "mcp-server-mas-sequential-thinking"
      ]
    }
  ]
}

Continue는 맵이 아닌 서버 오브젝트 배열 사용.

~/.config/zed/settings.json

{
  "context_servers": {
    "server-mas-sequential-thinking": {
      "command": {
        "path": "uvx",
        "args": [
          "mcp-server-mas-sequential-thinking"
        ]
      }
    }
  }
}

context_servers에 추가. 저장 시 Zed가 핫 리로드.

claude mcp add server-mas-sequential-thinking -- uvx mcp-server-mas-sequential-thinking

한 줄 명령. claude mcp list로 확인, claude mcp remove로 제거.

사용 사례

실전 활용법: mcp-server-mas-sequential-thinking

Stress-test a strategic decision with six perspectives

👤 Founders, PMs facing non-obvious tradeoffs ⏱ ~15 min intermediate

언제 쓸까: You're debating something hard ('raise now or bootstrap?', 'kill the feature?') and want sharper internal debate than single-agent brainstorming.

사전 조건

LLM API key (DeepSeek recommended for cost) — platform.deepseek.com

흐름

Frame the question

sequentialthinking: Should we raise a Series A now at $20M post to extend runway 24 months, or bootstrap further and raise at a likely higher valuation in 12 months? Here's our context: [facts]✓ 복사됨

→ Six-agent analysis + synthesis recommendation
Interrogate the synthesis

The synthesis recommends X. What's the weakest point in that reasoning? Have the critical agent push harder.✓ 복사됨

→ Sharper critique

결과: A well-rounded, written-down analysis you can share with co-founders.

함정

Token cost explosion on long context — Keep the framing tight — 200-500 words of context; agents multiply it 6x
Synthesis can average to mush — Pass the recommendation back through with decide between options A and B, don't hedge

Have the critical and emotional agents red-team your plan

👤 Anyone shipping a risky plan ⏱ ~10 min beginner

언제 쓸까: You drafted a launch plan and want the critical + emotional agents specifically to pressure-test it.

흐름

Submit the plan

sequentialthinking with focus on critical + emotional: Here's my launch plan [paste]. Find the three biggest risks and the three emotional reactions customers will have.✓ 복사됨

→ Two-perspective analysis

결과: Risks you hadn't noticed; emotional reactions to prepare for.

함정

Using this for simple yes/no questions wastes tokens — Reserve for multi-faceted decisions; use regular Claude for simple ones

Brainstorm with the creative agent grounded in real web data

👤 Content strategists, product people ⏱ ~15 min intermediate

언제 쓸까: You need ideas that aren't just from the model's training data — agents should search the web for current angles.

사전 조건

Exa API key — exa.ai — separate subscription

흐름

Enable Exa

Set EXA_API_KEY in env. Run sequentialthinking on: 'novel marketing angles for a local bakery in 2026'. Let creative + factual agents search the web.✓ 복사됨

→ Grounded ideas with citations

결과: Current, defensible creative angles.

함정

Exa search cost on top of LLM cost — Cap agent search budget via Exa rate limits

조합

다른 MCP와 조합해 10배 효율

server-mas-sequential-thinking + notion

Save the six-agent analysis as a decision doc

Run sequentialthinking on this strategic question, then create a Notion page in 'Decision Log' with all agent outputs and the synthesis.✓ 복사됨

도구

이 MCP가 노출하는 것

도구	입력	언제 호출	비용
sequentialthinking	prompt: str, perspectives?: str[], model?: str	Multi-faceted questions that benefit from different cognitive modes	5-10x single-agent tokens

비용 및 제한

운영 비용

API 쿼터: Bounded by your LLM provider quota
호출당 토큰: 5-10x a single-agent call — easily 20k-80k tokens per invocation
금액: DeepSeek is cheap (~$0.14/M input); with Anthropic Opus expect $1-3 per invocation
팁: Default to DeepSeek; only use Opus for the synthesis step via dual-model config.

보안

권한, 시크릿, 파급범위

자격 증명 저장: LLM_PROVIDER_API_KEY + optional EXA_API_KEY in env vars

데이터 외부 송신: Your prompts go to 6 agents' worth of LLM calls + optional Exa searches

High token usage — set daily spend caps with your LLM provider.
Not suitable for latency-sensitive contexts; each call is tens of seconds.

문제 해결

자주 발생하는 오류와 해결

Very slow / timeout

6 agents running sequentially = 30-60s. Increase client timeout to 120s+.

OPENAI_API_KEY / DEEPSEEK_API_KEY not found

Set the env var matching your chosen provider. Config file selects the provider.

Output inconsistent between runs

Stochastic by design — lower temperature or fix seed if your LLM supports it.

대안

mcp-server-mas-sequential-thinking 다른 것과 비교

대안	언제 쓰나	단점/장점
Native sequential-thinking MCP (official)	You want single-agent chain-of-thought without the multi-agent cost	Simpler, faster, cheaper; less diverse output
Direct use of Agno framework	You want to customize the agent roles	Not an MCP; code-first

mcp-server-mas-sequential-thinking

왜 쓰나요

핵심 기능

라이브 데모

실제 사용 모습

설치

클라이언트 선택

사용 사례

실전 활용법: mcp-server-mas-sequential-thinking

Stress-test a strategic decision with six perspectives

사전 조건

흐름

함정

Have the critical and emotional agents red-team your plan

흐름

함정

Brainstorm with the creative agent grounded in real web data

사전 조건

흐름

함정

조합

다른 MCP와 조합해 10배 효율

도구

이 MCP가 노출하는 것

비용 및 제한

운영 비용

보안

권한, 시크릿, 파급범위

문제 해결

자주 발생하는 오류와 해결

대안

mcp-server-mas-sequential-thinking 다른 것과 비교

더 보기

리소스