Local 5-agent career mentor drops with MCP

The project demonstrates a sophisticated implementation of Anthropic's Model Context Protocol (MCP) within a local-first agentic workflow, bypassing the privacy risks of cloud-based APIs.

• –Chained agent architecture allows for progressive context enrichment, resulting in highly personalized and smarter career reports as each agent learns from the previous one's output.
• –100% local execution using Ollama and Llama 3 addresses the privacy concerns of uploading sensitive resumes and personal data to external providers.
• –MCP integration standardizes the tool layer, enabling the system to be called directly from Claude Desktop or other MCP-compatible clients.
• –FAISS-based RAG ensures the agents remain grounded in the user's specific professional history and private knowledge base.
• –Technical implementation of JSON-RPC over stdio for MCP communication is a solid architectural choice for cross-process agent communication.