In this project, you’ll extend your Slack agent by connecting it to an MCP (Model Context Protocol) server. The server retrieves relevant context from a Qdrant knowledge base (prebuilt from a real course transcript) and sends a fully constructed RAG prompt to a local LLM. This separation of logic introduces clean, scalable architecture patterns.

You will:

• Build an MCP server that handles context retrieval and prompt construction
• Use LangChain RunnableSequence or LangGraph to define a RAG flow
• Query a prebuilt Qdrant collection containing your instructor’s transcript
• Generate high-quality answers with added context from retrieved documents
• Return answers to Slack via the original RTM agent
• Learn to separate agent input/output from processing logic using the Model Context Protocol