Type: url
Original: https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents
Status: generated
Summary
Anthropic’s Applied AI team presents context engineering — curating and maintaining the optimal set of tokens available to an LLM during inference — as the natural progression of prompt engineering for agents that operate over many turns. It argues that context must be treated as a finite resource subject to context rot and a limited attention budget, and sets out what a well-calibrated system prompt, tool set, and set of examples look like. For long-horizon work it describes just-in-time context retrieval alongside three techniques for working around the context window limit: compaction, structured note-taking, and sub-agent architectures.