Defines the workflow versus AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → distinction and argues for the simplest solution first, since agentic systems trade latency and cost for performance. Useful for deciding when an agent is warranted before adopting a framework.
Key takeaways · AI-distilled
Anthropic defines workflows as LLMs and tools orchestrated through predefined code paths, and agents as systems where the LLMA large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.Full definition → directs its own process and tool useA model's ability to call external functions — run code, search the web, edit files — instead of only generating text.Full definition →; workflows suit well-defined tasks, agents suit flexible decisions at scale.
For many applications, Anthropic says optimizing a single LLM call with retrieval and in-context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition → examples is enough, and building no agentic system at all may be the right choice.
Frameworks simplify setup but add abstraction that hides prompts and responses; Anthropic advises starting with direct LLM API calls, and says wrong assumptions about framework internals are a common source of customer error.
The basic building block is an augmented LLM with retrieval, tools and memory; Anthropic recommends tailoring these to the use case and giving the model a simple, well-documented interface to them.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
tool use — A model's ability to call external functions — run code, search the web, edit files — instead of only generating text.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
