📌Overview

Purpose: To enhance Claude Desktop's problem-solving capabilities using advanced reasoning methods, specifically through Beam Search and Monte Carlo Tree Search (MCTS).

Overview: The MCP Reasoner is an innovative reasoning framework designed to improve Claude Desktop's ability to tackle complex problems. With the introduction of experimental algorithms, it aims to integrate policy simulation with search methods, resulting in more efficient reasoning processes.

Key Features:

• Beam Search & MCTS: Offers two robust search strategies tailored for different problem complexities, allowing users to switch between straightforward (Beam Search) and complex (MCTS) scenarios.

• Path Tracking & Analysis: Continuously evaluates different reasoning paths and analyzes the effectiveness of each approach, providing insights into the reasoning process.

• MCP Protocol Compliance: Ensures adherence to the MCP protocol, integrating seamlessly with existing infrastructure for optimal performance.

---

MCP Reasoner

MCP Reasoner is a reasoning implementation for Claude Desktop that supports both Beam Search and Monte Carlo Tree Search (MCTS). It was developed to enhance Claude's complex problem-solving abilities.

Current Version

v2.0.0

What's New

• Added 2 Experimental Reasoning Algorithms:

  • mcts-002-alpha

    • Uses the A* Search Method along with an early alpha implementation of a Policy Simulation Layer
    • Includes early alpha versions of Adaptive Exploration Simulator & Outcome Based Reasoning Simulator
    • Note: These alpha simulators are not complete and subject to change

  • mcts-002alt-alpha

    • Uses the Bidirectional Search Method along with an early alpha implementation of a Policy Simulation Layer
    • Includes early alpha versions of Adaptive Exploration Simulator & Outcome Based Reasoning Simulator
    • Note: These alpha simulators are not complete and subject to change

What happened to mcts-001-alpha and mcts-001alt-alpha?
They were found to be ineffective and similar to the base mcts method, showing minimal improvements.

Why add Policy Simulation Layer now?
It is important to incorporate Policy and Search in tandem as most algorithms implement them together.

Previous Versions

v1.1.0

• Added model control over search parameters:
  • beamWidth: lets Claude adjust how many paths to track (1-10)
  • numSimulations: fine-tune MCTS simulation count (1-150)

Features

• Two search strategies:
  • Beam search (for straightforward problems)
  • MCTS with alpha variations (for complex problems)
• Tracks quality of different reasoning paths
• Maps Claude’s reasoning approaches
• Analyzes reasoning process
• Follows the MCP protocol

Installation

git clone https://github.com/frgmt0/mcp-reasoner.git

# Or clone the original:
git clone https://github.com/Jacck/mcp-reasoner.git

cd mcp-reasoner
npm install
npm run build

Configuration

Add to Claude Desktop config:

{
  "mcpServers": {
    "mcp-reasoner": {
      "command": "node",
      "args": ["path/to/mcp-reasoner/dist/index.js"]
    }
  }
}

Testing

More testing details will be added soon.

Benchmarks

Key benchmarks to test against include:

• MATH500
• GPQA-Diamond
• GMSK8
• Possibly Polyglot or SWE-Bench

Benchmarking data will be added soon.

License

This project is licensed under the MIT License - see the LICENSE file for details.