Math Tutor is an intelligent, agentic AI assistant designed to help users solve math problems using a combination of document retrieval, web search, and reasoning via LLMs. The system is modular, agent-based, and built using modern protocols like Model Context Protocol (MCP) and Agent to Agent (A2A) Protocol, making it extensible and production-ready.
- Retrieves relevant context from a vector database (AstraDB)
- Falls back to web search if vector search yields no results
- Uses LLMs for reasoning, extracting and solving math questions
- MCP-based microservices for modularity and scalability
- LangGraph orchestration for flexible agent workflows
- A2A-compatible responses with structured artifacts
- Automatic result saving (raw + cleaned formats) for analysis