Meta-Prompting

This MCP server implements the Meta-Prompting technique from the paper 'Meta-Prompting: Enhancing Language Models with Task-Agnostic Scaffolding', transforming a standard language model into a dynamic multi-agent system where the LM adopts two key roles: a Conductor that analyzes complex problems and breaks them into subtasks, and specialized Experts that execute each subtask. Built with Python using FastMCP and supporting multiple LLM providers (OpenAI, Anthropic, Google), the implementation simulates collaborative workflows within a single model through guided system prompts, allowing the model to reason, act, and self-critique for more robust solutions. The server provides tools for meta_model_prompt activation and expert_model consultation, with a fallback mechanism for clients that don't support ctx.sample(), making it valuable for complex problem-solving workflows, architectural decisions, and building AI assistants that benefit from structured multi-perspective reasoning without the overhead of managing separate model instances.