How to build a AI Agent
- Agentic AI is all about building an autonomous agent which can act on its own based on Environment Provided and if required also can bring Human in the loop.
- Examples:
- An Agent to automatically summarize the minutes of meeting (MOM)
- An Agent to respond to customer requests/issues raised.
- An Agent which helps in recruitment.
How to build
- This can be decomposed into two major components
- Brain: Choosing a Right LLM (model)
- Environment:
- Interacting with Brain -> Framework
- openai
- gemini
- llama
-
Environment: To interact with environment, llms started supporting which are function calls (tool call).
-
As AI Developers, all started building tools
- During November 2024, Claude Has released a paper which does nothing new but formulates a Standard for LLM => tool interaction called as Model Context Protocol
Some existing mcp servers which are already available
- Refer Here for mcp servers which are already available
