🦝 Tanuki Sequential Thought MCP

A sequential thinking framework for AI assistants that transforms unstructured thoughts into organized, executable tasks using your IDE’s built-in LLM capabilities.

🚀 Installation via Smithery

Install this MCP server instantly through Smithery:

1. Visit smithery.ai and search for “tanukimcp-thought”
2. Click “Install” and copy the generated configuration
3. Add the configuration to your MCP client (Claude Desktop, Cursor, etc.)
4. Restart your client - you’re ready to go!

Manual Installation

If you prefer manual installation, add this to your MCP configuration:

{
  "mcpServers": {
    "tanukimcp-thought": {
      "command": "npx",
      "args": ["@applejax2/tanukimcp-thought@latest"]
    }
  }
}

🧠 What is Sequential Thinking?

This MCP implements a structured approach to breaking down complex projects:

1. 🧠 Brain Dump → Transform scattered thoughts into organized tasks
2. 🔍 Find Next → Identify the logical next step to implement
3. 📋 Plan → Create detailed implementation strategy
4. 🛠️ Execute → Actually build/modify files in your workspace
5. ✅ Complete → Mark tasks as finished and track progress

🛠️ Available Tools

Core Sequential Workflow

• brain_dump_organize - Transform unstructured thoughts into structured todolist
• find_next_task - Identify the next logical task to implement
• plan_task_implementation - Create detailed implementation plan
• task_executor - Execute plans by creating/modifying files
• mark_task_complete - Mark tasks as complete in todolist

File Operations

• create_file - Create new files with content
• edit_file - Edit existing files (replace, append, prepend, insert)
• delete_file - Delete files safely
• move_file - Move files between locations
• copy_file - Copy files to new locations
• create_directory - Create directories
• list_directory - List directory contents
• delete_directory - Delete directories
• batch_operations - Execute multiple file operations

🎯 How to Use

Step 1: Organize Your Thoughts

Start any project by dumping your ideas:

Use brain_dump_organize with:
- project_description: "Personal Finance Tracker App"
- unstructured_thoughts: "Need login system, expense tracking, budget goals, charts, mobile responsive, dark mode, export data"
- workspace_root: "/path/to/your/project"

Step 2: Follow the Sequential Flow

1. Use find_next_task with your todolist file
2. Use plan_task_implementation for the identified task  
3. Use task_executor to implement the plan
4. Use mark_task_complete when done
5. Repeat until project complete!

Example: Building a Weather App

## Brain Dump
Use brain_dump_organize:
- project_description: "Desktop Weather App"
- unstructured_thoughts: "Fetch weather API, GUI interface, location search, 5-day forecast, weather icons, unit conversion, save favorites"
- workspace_root: "C:/Users/yourname/WeatherApp"

## Find & Execute Tasks
1. find_next_task → "Set up project structure"
2. plan_task_implementation → Creates detailed plan
3. task_executor → Builds the files and folders
4. mark_task_complete → Updates todolist

Repeat this cycle for each task!

🔑 Critical: Workspace Root Parameter

EVERY tool requires the workspace_root parameter - this must be the absolute path to where you want files created.

✅ Correct Usage:

workspace_root: "C:/Users/yourname/my-project"
workspace_root: "/home/user/projects/my-app"

❌ Incorrect Usage:

workspace_root: "."
workspace_root: "my-project"  
# Missing workspace_root parameter

For AI Assistants

When using these tools, always check the conversation context for the user’s current working directory and pass it as workspace_root. Look for:

• Terminal output showing current directory
• User statements like “I’m working in /path/to/project”
• File paths mentioned in conversation

🧠 IDE LLM Integration

This MCP exclusively uses your IDE’s built-in LLM capabilities:

• Claude Desktop - Uses Claude for intelligent task analysis
• Cursor - Leverages integrated Claude/GPT models
• VS Code - Works with Claude or Copilot extensions
• Other MCP Clients - Uses whatever LLM the client provides

No external dependencies - no local LLM installation required. The intelligence comes from your IDE’s LLM integration.

🎯 Example Workflows

Web Application Development

1. brain_dump_organize: "React todo app with authentication"
   - thoughts: "Login page, todo list, add/edit/delete todos, user profiles, responsive design"
   
2. find_next_task → "Set up React project structure"
3. plan_task_implementation → Creates package.json, folder structure plan
4. task_executor → Actually creates files and folders
5. Continue cycle for each feature...

API Development

1. brain_dump_organize: "REST API for blog platform"
   - thoughts: "User auth, CRUD posts, comments, file uploads, rate limiting"
   
2. Sequential implementation of each endpoint
3. Each task creates actual code files in your workspace

Desktop Application

1. brain_dump_organize: "Python GUI calculator app"
   - thoughts: "Basic operations, scientific mode, history, themes, keyboard shortcuts"
   
2. Build systematically from basic UI to advanced features
3. Each step creates real Python files you can run

🔧 Advanced Features

Batch Operations

Execute multiple file operations at once:

batch_operations:
- operations: [
    {type: "create_file", path: "src/main.py", content: "..."},
    {type: "create_directory", path: "tests"},
    {type: "copy_file", source: "template.js", destination: "src/app.js"}
  ]

Smart Task Dependencies

The find_next_task tool analyzes your todolist and suggests tasks based on:

• Logical dependencies (database before API routes)
• Complexity progression (simple features before complex ones)
• Current project state

🚨 Important Notes

1. Workspace Root is Required - Every tool needs the absolute path to your project
2. IDE LLM Only - This MCP uses your IDE’s LLM exclusively, no fallback logic
3. Real File Operations - Tools create/modify actual files in your workspace
4. Sequential Approach - Follow the workflow for best results, but tools work independently too

📄 License

MIT License - Build amazing things! 🚀

🔗 Links

• Smithery Installation
• MCP Documentation
• Report Issues