MCP-Chatbot

# MCP-Chatbot CLI-based MCP chatbot built with FastMCP Python SDK and Anthropic Claude API. Includes custom document read/write tools using Model Context Protocol architecture. A high-fidelity, CLI-based MCP chatbot designed to investigate the intersection of Agentic Tool-Use and Interaction Security. Built using the FastMCP Python SDK and Anthropic Claude API, this project serves as an empirical research tool for testing how frontier models manage persistent personas while executing sensitive read/write functions via the Model Context Protocol (MCP). Core Research Objectives: Functional Integrity: Evaluating the model's ability to maintain complex instructions while interacting with local file-system tools. Adversarial Robustness: Testing persona stability and safety guardrails during long-context, conversational interactions. Audit Workflows: Utilizing custom document tools to simulate real-world B2B security audits and data-handling scenarios.