# Erik Schluntz et al. - Building Effective Agents (Highlights)  ## Metadata **Review**:: [readwise.io](https://readwise.io/bookreview/47316209) **Source**:: #from/readwise #from/reader **Zettel**:: #zettel/fleeting **Status**:: #x **Authors**:: [[Erik Schluntz]], [[Barry Zhang]] **Full Title**:: Building Effective Agents **Category**:: #articles #readwise/articles **Category Icon**:: 📰 **URL**:: [www.anthropic.com](https://www.anthropic.com/research/building-effective-agents) **Host**:: [[www.anthropic.com]] **Highlighted**:: [[2024-12-31]] **Created**:: [[2025-01-19]] ## Highlights - Consistently, the most successful implementations weren't using complex frameworks or specialized libraries. Instead, they were building with simple, composable patterns. ([View Highlight](https://read.readwise.io/read/01jgfb0pwrhjr99ns3p8rvhyp7)) ^832154983 - **Agents**, on the other hand, are systems where LLMs dynamically direct their own processes and tool usage, maintaining control over how they accomplish tasks. ([View Highlight](https://read.readwise.io/read/01jgfb1tc4hn0qjrfr7k37wv4r)) ^832155171 - When building applications with LLMs, we recommend finding the simplest solution possible, and only increasing complexity when needed. ([View Highlight](https://read.readwise.io/read/01jgfb28v63jgsv5rb83d4dz6e)) ^832155198 - Agentic systems often trade latency and cost for better task performance, and you should consider when this tradeoff makes sense. ([View Highlight](https://read.readwise.io/read/01jgfb2pahjrk1kj2va3kqsxe8)) ^832155209 - When more complexity is warranted, workflows offer predictability and consistency for well-defined tasks, whereas agents are the better option when flexibility and model-driven decision-making are needed at scale. ([View Highlight](https://read.readwise.io/read/01jgfb3pqd84d7za5dtxyr7a3a)) ^832155230 - There are many frameworks that make agentic systems easier to implement, including: • [LangGraph](https://langchain-ai.github.io/langgraph/) from LangChain; • Amazon Bedrock's [AI Agent framework](https://aws.amazon.com/bedrock/agents/); • [Rivet](https://rivet.ironcladapp.com/), a drag and drop GUI LLM workflow builder; and • [Vellum](https://www.vellum.ai/), another GUI tool for building and testing complex workflows. ([View Highlight](https://read.readwise.io/read/01jgfb4aadgr1q2mg8tpyn7e1g)) ^832155244 - We suggest that developers start by using LLM APIs directly: many patterns can be implemented in a few lines of code. If you do use a framework, ensure you understand the underlying code. Incorrect assumptions about what's under the hood are a common source of customer error. ([View Highlight](https://read.readwise.io/read/01jgfb5jda3bv0a7by6csr22hr)) ^832155282 - The basic building block of agentic systems is an LLM enhanced with augmentations such as retrieval, tools, and memory. ([View Highlight](https://read.readwise.io/read/01jgfb89a7ne3zjagjpc5gfeex)) ^832155540 - LLMs can sometimes work simultaneously on a task and have their outputs aggregated programmatically ([View Highlight](https://read.readwise.io/read/01jgfc2mk8zpgs34snmwy7hd3v)) ^832157863 - In the orchestrator-workers workflow, a central LLM dynamically breaks down tasks, delegates them to worker LLMs, and synthesizes their results. ([View Highlight](https://read.readwise.io/read/01jgfc4rc7v9hy7xt10tbm1jsz)) ^832157924 - In the evaluator-optimizer workflow, one LLM call generates a response while another provides evaluation and feedback in a loop. ([View Highlight](https://read.readwise.io/read/01jgfc5gmv79wdnyyyd67er40s)) ^832157945 - Agents can handle sophisticated tasks, but their implementation is often straightforward. They are typically just LLMs using tools based on environmental feedback in a loop. ([View Highlight](https://read.readwise.io/read/01jgfc79k2v560wba80kedydth)) ^832158304