{
  "id": "agentic-context-engineering",
  "code": "PTL-0214",
  "term": "Agentic Context Engineering",
  "aliases": [
    "ACE"
  ],
  "category": "optimization",
  "definition": "Agentic Context Engineering (ACE) is a method for improving a language model application by treating its context, such as a system prompt or agent memory, as an evolving playbook that a loop of generation, reflection, and curation updates incrementally from execution feedback.",
  "description": "ACE adapts the inputs to a model rather than its weights. Its authors identify two failure modes of earlier context-adaptation methods, brevity bias, where summaries drop domain detail, and context collapse, where repeated rewriting erodes accumulated knowledge, and address them by adding and refining individual entries instead of rewriting the whole context. It can run offline to build a system prompt or online as agent memory, and learns from natural execution feedback without labeled data.",
  "example": null,
  "broader": [
    "prompt-optimization"
  ],
  "narrower": [],
  "related": [
    "context-engineering",
    "agent-memory",
    "reflexion",
    "self-refine"
  ],
  "introduced": 2025,
  "added": "2026-10-14",
  "sources": [
    {
      "title": "Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models",
      "authors": "Zhang et al.",
      "year": 2025,
      "url": "https://arxiv.org/abs/2510.04618"
    }
  ],
  "url": "https://protologue.com/t/agentic-context-engineering/",
  "citation": "Protologue. (2026). Agentic Context Engineering. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0214). https://protologue.com/t/agentic-context-engineering/"
}