# Sequential Thinking Multi-Agent System (MAS)

An MCP server that runs a structured thought through six specialist AI agents in parallel, each examining it from a different cognitive angle, before a synthesis agent produces one integrated answer.

This is an MCP server, not a standalone application: it runs as a background service that extends an MCP-compatible LLM client, such as Claude Desktop, with a single tool called sequentialthinking for structured multi-agent reasoning.

Every request follows the same fixed path: an initial synthesis, six specialist agents thinking in parallel, one each for factual, emotional, critical, optimistic, creative and meta-cognitive perspectives, and a final synthesis agent that integrates their output into one answer. Every specialist agent except synthesis can optionally use web research, and a complexity analyzer records diagnostic metadata about each request without currently changing which path it takes.

It is built with the Agno agent framework in Python, and configures two models per provider: an enhanced model used for synthesis and a separate model for the specialist agents.

Website: https://github.com/FradSer/mcp-server-mas-sequential-thinking