From Primitives to Production: How Anthropic Builds Agents
Anthropic defines agents by leaning on model intelligence with minimal core primitives in a loop.
“Agents in anthropic are actually very simply defined where essentially you want to give the model and lean in on model intelligence as much as possible.”
An Anthropic Applied AI engineer presented the company's internal agent philosophy at Databricks: agents are defined by giving models a small set of core primitives (code execution, file system, web search, to-do list) and trusting model intelligence to drive decisions, rather than relying on human-encoded deterministic logic. The talk traced the evolution from single-LLM calls to chained workflows to true agents, framing Claude Code as the canonical example. The insight matters because it reveals Anthropic's production stance — maximize model autonomy, minimize orchestration scaffolding — which directly shapes how enterprise customers should design their own agent systems.