This MCP server implements a Sequential Thinking Multi-Agent System (MAS) built with the Agno framework and served via MCP. It provides a dedicated sequentialthinking tool that orchestrates six specialized agents—Factual, Emotional, Critical, Optimistic, Creative, and Synthesis—to analyze a prompt from multiple cognitive angles, then synthesize a coherent, user-friendly answer. The system uses a fixed processing strategy called full_exploration, where all agents run in parallel where appropriate and the Synthesis agent integrates their perspectives to deliver the final response. ExaTools-enabled research (via the Factual, Critical, Optimistic, and Creative agents) can pull current facts, counterexamples, success stories, and cross-industry insights when an EXA_API_KEY is provided; otherwise the reasoning is purely internal. Use the server as a drop-in enhancement to LLM clients to obtain more structured, multi-perspective reasoning and actionable insights.

How to use the capabilities:

• Start the MCP server and connect your LLM client to the Sequential Thinking tool (sequentialthinking) exposed by the server.
• Submit a thought or problem statement; the system will route the task through the six agents, collecting facts, emotional intuitions, risk assessments, optimistic opportunities, creative alternatives, and finally a synthesized, actionable answer.
• If you provide EXA_API_KEY, enable ExaTools-driven web research to augment the agents’ outputs with up-to-date data and sources. Without the key, the system relies on internal reasoning and known knowledge.
• Review the final synthesis, which includes integrated perspectives and any recommended next steps or decisions.

Prerequisites:

• Python 3.10 or newer installed on your system
• Basic virtual environment tooling (optional but recommended)
• Internet access for package installation

1. Create and activate a virtual environment (recommended): python -m venv mcp-env

   Windows

   mcp-env\Scripts\activate.bat

   macOS/Linux

   source mcp-env/bin/activate

2. Install FastMCP and Agno (and any other dependencies): pip install fastmcp agno

3. Install or ensure the MAS SEQUENTIAL THINKING server package is available in your environment

   If published as a package, install it similarly to any MCP server package

   Example (if available):

   pip install fradser-mcp-server-mas-sequential-thinking

4. Run the MCP server using the Python module entry point specified in the configuration: python -m fastmcp.server

5. Optional: set environment variables before starting the server:

   • EXA_API_KEY: Your ExaTools API key for research capabilities

6. Connect your MCP client to the server endpoint and start issuing prompts to the sequentialthinking tool.

Tips and common considerations:

• EXA_API_KEY enables ExaTools-based web research for several agents (Factual, Critical, Optimistic, Creative). The system remains functional without it, but results may be less data-rich.
• The AI routing uses a fixed strategy (full_exploration); no legacy or alternate routing modes are active.
• Time allocations per agent are defined in the architecture: Factual (120s), Emotional (30s), Critical (120s), Optimistic (120s), Creative (240s), and Synthesis (60s).
• To customize behavior, consider adjusting exposure to ExaTools (toggle via environment) and enabling logging for observability (structured Python logging recommended).
• This server is designed as an MCP back-end to augment LLM clients (e.g., Claude Desktop) with multi-agent reasoning capabilities rather than a standalone interactive app.