protologue

Agentic Context Engineering

Also called ACE.

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.

Sources

  1. Zhang et al. (2025). Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models.

Cite this entry

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/

BibTeX
@misc{protologue_agentic_context_engineering,
  title = {Agentic Context Engineering},
  author = {{Protologue}},
  year = {2026},
  howpublished = {Protologue: A Taxonomy of Prompting and LLM Techniques, v1.4.0},
  note = {Entry PTL-0214},
  url = {https://protologue.com/t/agentic-context-engineering/}
}

Markdown JSON