Agent framework (LangChain, LangGraph and others)
A library for building AI agents and LLM workflows: model calls, tools, state, retries and tracing, so you do not write the loop yourself.
LangChain and LangGraph, LlamaIndex, CrewAI and the model vendors’ own agent SDKs are common choices. They speed up prototypes and give checkpoints, human approval steps and tracing out of the box. The cost is abstraction you have to debug through and upgrades that change APIs. An agent loop itself is a few dozen lines; the hard parts are tools, limits and evals, which no framework writes for you.
Related terms
- Agentic loop: The cycle of model call, tool call and result that an agent repeats until it stops.
- Agent harness (harness engineering): Everything around the model that makes it an agent: the loop, tools, permissions, memory, context management.
- Durable execution: Running a long process so its progress is saved after every step, letting it survive crashes and pause for hours or days, then resume.
- Agent graph (state graph): Modeling an agent as explicit nodes, edges and shared state, with checkpoints, so each step can be paused, resumed, inspected and tested.
