Fradser MCP Server Mas Sequential Thinking

🚀 Sequential Thinking Tool Documentation

This documentation describes a tool named sequentialthinking that implements step-by-step reasoning through a Multi-Agent System (MAS). This tool enables Large Language Models (LLMs) to interact with a server endpoint to complete complex step-by-step thinking tasks.

🚀 Quick Start

Features

The sequentialthinking tool allows LLMs to interact with a server endpoint, enabling them to perform complex step-by-step thinking tasks through a Multi-Agent System (MAS).

📦 Installation

1. Clone the repository.
2. Set up a virtual environment.
3. Install the dependencies:

# Using uv
uv pip install -e ".[dev]"
# Using pip
pip install -e ".[dev]"

💻 Usage Examples

Basic Usage

The process of an LLM calling the tool is as follows:

1. The LLM initiates thinking using the sequential-thinking-starter prompt.
2. The LLM calls the sequentialthinking tool with an initial thought number of 1, content such as "Planning analysis...", and specifies the total number of steps and whether the next step is needed.
3. The server endpoint processes the thought, and the MAS synthesizes a response to provide guidance for the next step.
4. The LLM receives the synthesized response and conducts the next step of thinking based on the guidance.
5. Repeat the above process until all thinking steps are completed.

📚 Documentation

Sequential Thinking Tool Parameters

The tool accepts a ThoughtData parameter that conforms to the Pydantic model:

{
    "thought": str,              # Current thought content
    "thoughtNumber": int,        # Serial number (>=1)
    "totalThoughts": int,        # Estimated total number of steps (>=1, recommended >=5)
    "nextThoughtNeeded": bool,   # Is the next step needed?
    "isRevision": bool = False,  # Is it a revision of the previous step?
    "revisesThought": Optional[int] = None, # If it's a revision, specify the thought number to be revised
    "branchFromThought": Optional[int] = None, # If it's a branch, specify the thought number of the source
    "branchId": Optional[str] = None, # Unique identifier for the branch
    "needsMoreThoughts": bool = False # Signal if the estimated total number of steps is too low?
}

Tool Response Format

The tool returns data in the following JSON format:

{
  "processedThoughtNumber": int,    # Processed thought number
  "estimatedTotalThoughts": int,     # Estimated total number of steps
  "nextThoughtNeeded": bool,         # Is the next step needed?
  "coordinatorResponse": "..."       # Synthesized output, including analysis, findings, and guidance for the next step...
}

🔧 Technical Details

Logging

• The log file is located at ~/.sequential_thinking/logs/sequential_thinking.log
• Uses Python's standard logging module
• Includes rotating file handling (10MB limit, 5 backups) and console output (INFO level)
• Records timestamps, log levels, logger names, and messages

📄 License

This tool is licensed under the MIT License.