Offers API integration for Node.js applications to utilize Smart-Thinking's reasoning capabilities programmatically, with functions for thought processing and verification.
Available as an npm package (smart-thinking-mcp) for easy installation and integration into JavaScript/TypeScript projects.
Built with TypeScript 5.1.6, providing type-safe integration for applications that want to incorporate Smart-Thinking's reasoning capabilities.
Smart-Thinking
Overview
Smart-Thinking is a sophisticated MCP (Model Context Protocol) server that provides a multi-dimensional, adaptive, and self-verifying reasoning framework for AI assistants like Claude. Unlike linear reasoning approaches, Smart-Thinking uses a graph-based architecture that allows for complex connections between thoughts, providing a more nuanced reasoning capability that is closer to human cognition.
Smart-Thinking is fully compatible with all platforms (Windows, macOS, Linux) and integrates seamlessly with many MCP clients, including Claude Desktop, Cline, Windsurf, and other MCP-compatible applications.
Key Features
Advanced cognitive architecture
- Multi-dimensional thought graph replacing traditional linear sequences
- Dynamic and self-adaptive estimation of the number of thoughts needed
- Rich and nuanced types of connections between thoughts (supports, contradicts, refines, etc.)
- Contextual thinking metrics (confidence, relevance, quality)
Robust verification system
- Automatic fact checking and calculations with 8 different statuses
- Detection of contradictions and uncertainties in reasoning
- Reliability scores tailored to each type of verification
- Generation of context-appropriate certainty summaries
Advanced Features
- Persistent memory for previous sessions
- Self-learning mechanisms to improve reasoning
- User-friendly personalization for relevant reasoning
- Interactive visualizations of the reasoning process
- Multi-agent collaboration for teamwork
- Seamless integration with the MCP tool ecosystem
Cross-platform compatibility
- Works identically on Windows, macOS, and Linux
- Compatible with NVM (Node Version Manager) on all platforms
- Automatic management of file paths according to the platform
- Simplified configuration for each environment
- Automatic resolution of OS-specific problems
Facility
Option 1: Global Installation (Recommended)
Option 2: Installation via Smithery
To install Smart-Thinking automatically via Smithery :
Option 3: Use via npx (without installation)
Option 4: Install from source code
Configuration with Claude Desktop
Setup on macOS
Add this configuration to your ~/Library/Application Support/Claude/claude_desktop_config.json
file:
OR via npx:
Configuration on Windows
Add this configuration to your %APPDATA%\Claude\claude_desktop_config.json
file:
OR with the full path to Node.js (recommended for Windows):
Important :
- Replace
VotreNom
with your Windows username. - Use forward slashes (
/
) in Windows paths, even if the OS uses backslashes (\
).
For detailed installation instructions, see the Installation Guide .
Cross-platform file system
Smart-Thinking implements advanced file system management compatible with all platforms:
File System Features
- Automatic path normalization : Seamless conversion between Windows (
\
) and Unix (/
) path separators - Built-in platform detection : Automatic adaptation according to the OS (Windows, macOS, Linux)
- Special Path Handling : Support for Windows UNC paths, WSL paths and paths with spaces
- Self-configured data directory : Automatic creation and management of the data directory
- Intelligent Fallback : Automatic creation of an alternative directory in case of access problems
- Platform-specific configuration paths : Correct location of configuration files depending on the OS
Configuring the data directory
By default, Smart-Thinking creates and uses a data
folder in its working directory. You can also specify a custom directory with the SMART_THINKING_DATA_DIR
environment variable:
Support for NVM (Node Version Manager)
Smart-Thinking automatically detects if Node.js is installed via NVM and adapts the paths accordingly, providing perfect compatibility on all platforms, including Windows with NVM.
Use
Main settings
Setting | Kind | Description |
---|---|---|
thought | string | Content of the thought to be analyzed (mandatory) |
thoughtType | string | Type of thinking ('regular', 'revision', 'meta', 'hypothesis', 'conclusion') |
connections | array | Connections to other thoughts |
requestVerification | boolean | Enable automatic verification |
containsCalculations | boolean | Indicate whether the thought contains calculations |
generateVisualization | boolean | Generate a thought graph visualization |
suggestTools | boolean | Suggest relevant MCP tools |
sessionId | string | Session ID for persistence |
Examples of use
Simple reasoning
With automatic verification
With visualization
Collaborative analysis
Verification system
Smart-Thinking incorporates a sophisticated verification system that evaluates the reliability of information and calculations. It supports 8 distinct verification statuses:
Status | Description |
---|---|
verified | Information verified with confidence |
partially_verified | Information partially verified |
unverified | Unverified information |
contradicted | Information contradicted by other sources |
inconclusive | Verification inconclusive |
absence_of_information | Lack of information on the subject |
uncertain | Uncertain information due to contradictions |
contradictory | Inherently contradictory information |
The system calculates a reliability score that combines these statuses with other metrics like trust, relevance, and quality.
Calculating the reliability score
The reliability score is calculated using a formula that balances:
- Basic metrics (trust, relevance, quality)
- The verification status
- The results of the calculation verification (if present)
- History of previous scores (for temporal smoothing)
Thresholds and scores have been optimized through simulation to ensure maximum consistency.
Visualization
Smart-Thinking offers several types of thought graph visualizations:
- Standard Graph : Standard arrangement of the thought network
- Chronological : Temporal organization of thoughts
- Theme : Clusters by similar themes
- Hierarchical : Tree structure
- Force : Arrangement based on the forces of attraction/repulsion
- Radial : Concentric circles around a central thought
Visualizations can be filtered according to:
- Types of thoughts
- Connection types
- Metric thresholds
- Text search
- Date ranges
Data persistence
Smart-Thinking implements a robust data persistence system that stores:
- Thought graphs by session
- Verification results for future reuse
- Metrics calculated for analysis and improvement
- User preferences for personalization
Data is stored in a cross-platform compatible manner in structured JSON files on the file system, ensuring persistence across sessions.
Compatibility with MCP clients
Smart-Thinking is compatible with many MCP clients, including:
- Claude Desktop App : Full support for tools, resources and prompts
- Cline : Support for tools and resources
- Continue : Full support for all MCP features
- 5th : Support for tools
- Cursor : Support for tools
- Windsurf Editor : Support for AI Flow tools
- And more...
Comparison with Sequential-Thinking
Functionality | Sequential Thinking | Smart-Thinking |
---|---|---|
Thought structure | Linear | Multi-dimensional (graph) |
Connection types | Boundaries | Rich and nuanced (16+ types) |
Adaptation | Static | Dynamic and contextual |
Verification | Basic | Advanced (8 different statuses) |
Visualization | Simple | Interactive and configurable |
Memory | Temporary | Persistent with vectorization |
Collaboration | No | Yes (multi-agent) |
Personalization | Limited | User-adaptive |
Self-learning | No | Yes |
Metrics | Basics | Contextual and multi-factor |
Platform compatibility | Limited | Complete (Windows, macOS, Linux) |
API and integration
Smart-Thinking can be integrated into other Node.js applications:
Troubleshooting
Quick Operation Check
To verify that Smart-Thinking is working correctly:
You should see the server starting message.
Consult the logs
If you encounter any difficulties, check the Claude Desktop logs:
- On macOS:
~/Library/Logs/Claude/mcp*.log
- On Windows:
%USERPROFILE%\AppData\Local\Claude\logs\mcp*.log
- On Linux:
~/.local/share/Claude/logs/mcp*.log
Common Problems and Solutions
Issue | Solution |
---|---|
Smart-Thinking does not appear in Claude | Check the paths in the configuration, restart Claude Desktop |
File access errors | Check the permissions of the data directory |
Connection problems on Windows | Use full and absolute paths with forward slashes |
Errors with NVM on Windows | Use the full path to the active version of Node.js |
For Windows and NVM specific issues, see the dedicated section in the Installation Guide .
Contribution
Contributions are welcome! Check out our Contribution Guide for more information.
License
You must be authenticated.
local-only server
The server can only run on the client's local machine because it depends on local resources.
Tools
A sophisticated MCP server that provides a multi-dimensional, adaptive reasoning framework for AI assistants, replacing linear reasoning with a graph-based architecture for more nuanced cognitive processes.
- Overview
- Key Features
- Facility
- Configuration with Claude Desktop
- Cross-platform file system
- Use
- Verification system
- Visualization
- Data persistence
- Compatibility with MCP clients
- Comparison with Sequential-Thinking
- API and integration
- Troubleshooting
- Contribution
- License
Related Resources
Related MCP Servers
- AsecurityFlicenseAqualityProvides reasoning content to MCP-enabled AI clients by interfacing with Deepseek's API or a local Ollama server, enabling focused reasoning and thought process visualization.Last updated -15424JavaScript
- AsecurityAlicenseAqualityA Model Context Protocol server that enables AI assistants to interact with Linear project management systems, allowing users to retrieve, create, and update issues, projects, and teams through natural language.Last updated -32805TypeScriptMIT License
- -securityAlicense-qualityA minimal MCP Server that provides Claude AI models with the 'think' tool capability, enabling better performance on complex reasoning tasks by allowing the model to pause during response generation for additional thinking steps.Last updated -5251TypeScriptMIT License
- AsecurityAlicenseAqualityAn MCP server that supercharges AI assistants with powerful tools for software development, enabling research, planning, code generation, and project scaffolding through natural language interaction.Last updated -1112TypeScriptMIT License