Anthropic：2024智能体（Agents）高效构建指南（英文版）（16页）.pdf

1、Engineering at AnthropicBuilding effective agentsWeve worked with dozens of teams building LLM agents acrossindustries.Consistently,the most successful implementations use simple,composable patterns rather than complex frameworks.Published Dec 19,2024Over the past year,weve worked with dozens of tea

2、ms building large languagemodel(LLM)agents across industries.Consistently,the most successfulimplementations werent using complex frameworks or specialized libraries.Instead,they were building with simple,composable patterns.In this post,we share what weve learned from working with our customersand

3、building agents ourselves,and give practical advice for developers onbuilding effective agents.What are agents?Agent can be defined in several ways.Some customers define agents as fullyautonomous systems that operate independently over extended periods,usingvarious tools to accomplish complex tasks.

4、Others use the term to describemore prescriptive implementations that follow predefined workflows.AtAnthropic,we categorize all these variations as agentic systems,but draw animportant architectural distinction between workflows and agents:Workflows are systems where LLMs and tools are orchestrated

5、throughpredefined code paths.Agents,on the other hand,are systems where LLMs dynamically directtheir own processes and tool usage,maintaining control over how theyaccomplish tasks.Below,we will explore both types of agentic systems in detail.In Appendix 1(“Agents in Practice”),we describe two domain

6、s where customers have foundparticular value in using these kinds of systems.When(and when not)to use agentsWhen building applications with LLMs,we recommend finding the simplestsolution possible,and only increasing complexity when needed.This mightmean not building agentic systems at all.Agentic sy