Imagine that every problem in your application could be solved with an LLM-based agent. It sounds like a promise of a new generation of frameworks... or the beginning of an unnecessarily complex architecture. Where does real value end and hype begin?
Agent frameworks are entering the world of software engineering with increasing confidence. LangChain and LangGraph have become two of the most frequently mentioned names in this ecosystem - sparking both excitement and a fair share of doubt. Do they truly simplify the development of intelligent systems? Or do they merely shift complexity to new layers?
In this presentation, we will take a practical look at the LangChain and LangGraph ecosystem. We will explain what agent frameworks are and the problems they aim to solve. We will discuss the role of both tools, the key differences between them, as well as typical use cases, limitations, and the costs introduced by additional abstraction.
Instead of marketing promises, we will focus on something tangible: a simple example of an SRE agent that, based on a brief incident report, helps with the initial diagnosis of an issue in a microservices-based application. Not as a production-ready solution, but as an illustration of what these tools can actually do - and what should not be expected from them.
If you are wondering whether LangChain or LangGraph fit your use cases, want to understand their real value, and avoid costly technological experiments, this talk will help you make an informed decision before investing your time and energy in their implementation.
