MCTS MCP Server

Example Use Cases

• 1

  Perform deep, iterative reasoning and exploration over a question or text using Monte Carlo Tree Search to surface diverse analytical pathways.

• 2

  Run multi-LLM experiments or compare runs across providers/models and generate structured reports, insights, and suggested improvements.

• 3

  Integration with Claude Desktop and stateful, resumable analyses with persistent run data and configurable MCTS parameters.

Description

This open-source MCP server provides a Bayesian MCTS engine and a suite of analysis tools that let an assistant (e.g., Claude) run multi-iteration, multi-simulation reasoning and syntheses. It supports state persistence, approach taxonomy, surprise detection, and configurable selection strategies (Thompson sampling or UCT). The server integrates with multiple LLM providers (Ollama, OpenAI, Anthropic, Google Gemini) via API keys and can be wired into Claude Desktop for interactive use. Results, metrics and reports are stored per run and can be inspected, compared, and exported.