Gen-AI Developer Classroom notes 29/Nov/2025

Application Types

• Generative AI: Generating content based on training it has undergone.

• Application Types

  • RAG (Retrieval Augmented Generation): Interacting with external/private data + LLM to generate a response
    
  • Agents: Interacting with external/private data + LLM and it should be performing activities. Agents can be interactive or non interactive.
    • Create a github issue
    • Backup the database
    • …

Agents

• Agents need to perform actions,
• Models like Open AI, Gemini started supporting tool calls
• To build agents we will be using frameworks like Langgraph, SemanticKernel.
• Each of this framework has a different way of designing a tool.
• Lets assume we are building a tool for github, For different llms i need to build different tools and for different framework a different
• Then Claude has launched a paper called mcp
• MCP (Model context protocol) is type-c charger of tools

MCP

• MCP has two major components

  • mcp-client
  • mcp-server
    

• Vendors are give mcp servers

  • Github
  • JIRA

• Community

• What we need to learn

  • Building & using mcp servers
  • integrating mcp servers with
    • your code
    • other tools
  • On june 2025, mcp started added authorizations