Source: Building effective agents (Anthropic)
Anthropic's guide to agentic systems: prefer the simplest solution, distinguish workflows from agents, and start with raw LLM APIs before adopting a framework.
Contents
- What: engineering blog post
- Author/publisher: Anthropic (Erik S. and Barry Zhang)
- Date: 2024-12-19 (the page now carries a note that the tooling landscape has changed since December 2024 and points to Claude Managed Agents)
- Original: anthropic.com/engineering/building-effective-agents
Summary
The post separates workflows (LLMs and tools orchestrated through predefined code paths) from agents (LLMs directing their own process and tool use), catalogues common workflow patterns, and recommends adding agentic complexity only when it measurably helps. It includes a short section on when and how to use frameworks.
Key claims
- Find the simplest solution first; for many applications, optimizing single LLM calls with retrieval and in-context examples is enough.
- Agentic systems trade latency and cost for task performance.
- Frameworks simplify low-level work (calling LLMs, defining and parsing tools, chaining calls), but extra abstraction can hide the underlying prompts and responses, making debugging harder, and makes it tempting to add complexity.
- Start with LLM APIs directly, since many patterns take a few lines of code; if you adopt a framework, understand the code underneath, because wrong assumptions about it are a common source of customer error.