think-mcp
Provides structured thinking tools including mental models, design patterns, debugging approaches, decision frameworks, and multi-persona reasoning to enhance AI assistant problem-solving capabilities.
README
think-mcp
Structured thinking tools for AI assistants. Provides mental models, debugging approaches, decision frameworks, and more via the Model Context Protocol.
Quick Start
Installation
npm install -g think-mcp
Claude Desktop Configuration
Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"think-mcp": {
"command": "npx",
"args": ["think-mcp"]
}
}
}
Tools
| Tool | Description | Use When |
|---|---|---|
trace |
Step-by-step thought tracing | Breaking down complex problems |
model |
Mental models (first principles, pareto, etc.) | Applying structured thinking frameworks |
pattern |
Software design patterns | Architectural decisions |
paradigm |
Programming paradigms | Choosing coding approaches |
debug |
Debugging approaches | Systematic troubleshooting |
council |
Multi-persona deliberation | Getting diverse perspectives |
decide |
Decision analysis frameworks | Making important choices |
reflect |
Metacognitive monitoring | Assessing knowledge boundaries |
hypothesis |
Scientific method | Testing ideas systematically |
debate |
Dialectical reasoning | Exploring arguments |
map |
Visual/spatial reasoning | Diagramming concepts |
Examples
Using trace for complex problem-solving
"Use the trace tool to think through how to refactor this authentication system"
Using model with first principles
"Apply first_principles thinking using the model tool to analyze why our build is slow"
Using council for multiple perspectives
"Use council to get security expert, performance engineer, and UX designer perspectives on this API design"
Using debug for systematic troubleshooting
"Use the debug tool with binary_search approach to find the source of this memory leak"
Using decide for structured decisions
"Use decide with weighted-criteria analysis to choose between PostgreSQL and MongoDB"
Tool Details
trace (Sequential Thinking)
Dynamic problem-solving through structured thoughts with revision and branching support.
- Break down complex problems into steps
- Revise previous thoughts as understanding evolves
- Branch into alternative approaches
- Track progress across multiple reasoning steps
model (Mental Models)
Apply proven mental models to problems:
first_principles- Break down to fundamentalsopportunity_cost- Analyze trade-offserror_propagation- Understand error chainsrubber_duck- Explain to clarifypareto_principle- Focus on high-impact factorsoccams_razor- Prefer simpler explanations
pattern (Design Patterns)
Software architecture patterns:
modular_architecture- Component separationapi_integration- External service patternsstate_management- State handling approachesasync_processing- Asynchronous patternsscalability- Growth considerationssecurity- Security best practicesagentic_design- AI agent patterns
paradigm (Programming Paradigms)
Programming approach selection:
imperative,procedural,object_orientedfunctional,declarative,logicevent_driven,aspect_orientedconcurrent,reactive
debug (Debugging Approaches)
Systematic debugging methods:
binary_search- Divide and narrowreverse_engineering- Work backwardsdivide_conquer- Isolate componentsbacktracking- Trace executioncause_elimination- Rule out causesprogram_slicing- Focus on relevant code
council (Collaborative Reasoning)
Multi-persona problem-solving with diverse expertise, structured debate, and consensus building.
decide (Decision Framework)
Structured decision analysis:
pros-cons- Simple comparisonweighted-criteria- Multi-factor scoringdecision-tree- Branching outcomesexpected-value- Probability-weightedscenario-analysis- Future scenarios
reflect (Metacognitive Monitoring)
Self-awareness about knowledge boundaries, claim certainty, and reasoning biases.
hypothesis (Scientific Method)
Formal scientific reasoning with hypothesis testing, variable identification, and evidence evaluation.
debate (Structured Argumentation)
Dialectical reasoning with thesis-antithesis-synthesis and argument strength analysis.
map (Visual Reasoning)
Visual thinking with diagrams, graphs, flowcharts, concept maps, and state diagrams.
Remote Access (Vercel Deployment)
Think-MCP can be deployed to Vercel for remote access from Claude Desktop and ChatGPT without local installation.
Deploy to Vercel
Or deploy manually:
# Install Vercel CLI
npm i -g vercel
# Deploy
vercel
Connect Claude Desktop (Remote)
Add to your claude_desktop_config.json:
{
"mcpServers": {
"think-mcp": {
"command": "npx",
"args": ["mcp-remote", "https://your-deployment.vercel.app/api/mcp"]
}
}
}
Connect ChatGPT
- Go to Settings → Connectors → Advanced → Developer Mode
- Add the URL:
https://your-deployment.vercel.app/api/mcp
Development
# Clone the repository
git clone https://github.com/chirag127/think-mcp.git
cd think-mcp
# Install dependencies
npm install
# Build
npm run build
# Run tests
npm test
# Development mode (watch)
npm run dev
# Web development (Vercel)
cd web && npm install
npm run dev:web
Project Structure
think-mcp/
├── dist/ # Compiled JavaScript files (npm package)
├── src/ # STDIO server source (for npx/local usage)
│ ├── models/ # Data interfaces
│ ├── tools/ # Tool implementations
│ │ ├── traceServer.ts
│ │ ├── modelServer.ts
│ │ ├── patternServer.ts
│ │ ├── paradigmServer.ts
│ │ ├── debugServer.ts
│ │ ├── councilServer.ts
│ │ ├── decideServer.ts
│ │ ├── reflectServer.ts
│ │ ├── hypothesisServer.ts
│ │ ├── debateServer.ts
│ │ └── mapServer.ts
│ ├── toolNames.ts # Tool name constants
│ └── index.ts # Main server entry point
├── web/ # Next.js app for Vercel deployment
│ ├── app/
│ │ ├── layout.tsx
│ │ ├── page.tsx # Landing page
│ │ └── api/
│ │ └── [transport]/
│ │ └── route.ts # MCP endpoint
│ └── lib/
│ ├── mcp-tools.ts # Tool registration
│ └── tools/ # Tool definitions
├── vercel.json # Vercel deployment config
├── package.json
├── tsconfig.json
└── README.md
Tech Stack
- TypeScript
- Node.js 18+
- Model Context Protocol SDK 1.25.1
- Zod 3.25+ (validation)
- Vitest (testing)
Author
Chirag Singhal (@chirag127)
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。