MCTS MCP Server

Instructions

Implementation Reference

• src/mcts_mcp_server/server.py:356-409 (handler)

  The primary handler function for the 'get_synthesis' tool. It generates a final comprehensive synthesis using the LLM based on the best MCTS analysis, checks prerequisites, and returns structured results or errors.

  async def get_synthesis() -> dict[str, Any]:
      """
      Generate a final synthesis of the MCTS results.
  
      Creates a comprehensive summary that synthesizes the key insights from the best
      analysis found during the MCTS search process.
  
      Returns:
          Dict containing the synthesis text, best score, and metadata
  
      Raises:
          Exception: If synthesis generation fails or MCTS hasn't been run yet
      """
      if not server_state["initialized"]:
          return {"error": "MCTS not initialized. Call initialize_mcts first.", "status": "error"}
  
      if server_state["best_score"] == 0.0:
          return {"error": "No analysis completed yet. Run run_mcts_search first.", "status": "error"}
  
      try:
          question = server_state["current_question"]
          best_analysis = server_state["best_analysis"]
          best_score = server_state["best_score"]
  
          synthesis_prompt = f"""Create a comprehensive synthesis based on this MCTS analysis:
  
  Original Question: {question}
  
  Best Analysis Found (Score: {best_score}/10):
  {best_analysis}
  
  Provide a final synthesis that:
  1. Summarizes the key insights
  2. Highlights the most important findings
  3. Offers actionable conclusions
  4. Explains why this approach is valuable
  
  Make it clear, comprehensive, and practical."""
  
          synthesis = await call_llm(synthesis_prompt)
  
          return {
              "synthesis": synthesis,
              "best_score": best_score,
              "iterations_completed": server_state["iterations_completed"],
              "question": question,
              "provider": server_state["provider"],
              "model": server_state["model"],
              "status": "success"
          }
  
      except Exception as e:
          logger.error(f"Error generating synthesis: {e}")
          return {"error": f"Synthesis failed: {e!s}", "status": "error"}

• src/mcts_mcp_server/server.py:110-114 (registration)

  Registration of the 'get_synthesis' tool in the @server.list_tools() handler, defining its name, description, and empty input schema (no parameters required).

  types.Tool(
      name="get_synthesis",
      description="Generate a final synthesis of the MCTS results",
      inputSchema={"type": "object", "properties": {}}
  ),

• src/mcts_mcp_server/server.py:113-113 (schema)

  Input schema for the 'get_synthesis' tool, which is an empty object indicating no input parameters are required.

  inputSchema={"type": "object", "properties": {}}

• src/mcts_mcp_server/server.py:154-155 (registration)

  Dispatch logic in the @server.call_tool() handler that routes calls to the get_synthesis function.

  elif name == "get_synthesis":
      result = await get_synthesis()