A Language for Describing Agentic LLM Contexts

2026-05-05 · Source: cs.MA updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Robotics & Autonomous Systems · Depth: Advanced, extended

Summary

The Agentic Context Description Language (ACDL) is introduced as a formal, human- and machine-readable language for precisely specifying the structure and dynamic evolution of Large Language Model (LLM) input contexts within agentic systems. Developed by Noga Peleg Pelc, Gal A. Kaminka, and Yoav Goldberg, ACDL addresses the current lack of a standard for communicating how LLM prompts are constructed and change across multi-turn interactions, which typically relies on informal prose or code inspection. ACDL provides constructs for defining role message sequences, dynamic content, time-indexed references, and conditional or iterative structures, abstracting away specific content to focus on the architectural blueprint of a prompt. The language supports visualization, with tooling including a web editor, VS Code extension, and Claude Code skill available at www.acdlang.org. It aims to improve communication, comparison, and reproducibility of LLM agent designs, demonstrating its utility by documenting existing systems like ReAct, OpenCode, OpenClaw, and the Gemini Plays Pokémon agent.

Key takeaway

For AI Architects and NLP Engineers designing or analyzing LLM-based agentic systems, adopting ACDL can significantly enhance clarity and precision. This language provides a standardized way to describe complex prompt structures and their temporal evolution, making it easier to communicate design choices, compare different agent implementations, and reproduce research findings. Consider integrating ACDL into your documentation and design workflows to foster more rigorous and transparent development of agentic LLM systems.

Key insights

ACDL standardizes describing LLM agent context structure and evolution, improving communication and reproducibility.

Principles

• Context structure directly shapes agent behavior.
• Formal languages enable rigorous analysis.
• Abstraction aids clear communication.

Method

ACDL specifies context as a sequence of role messages containing information pieces, categorized by source (constants, system state, environment state, LLM responses, functions). It uses time-indexing, control flow (ForEach, If/Else, Switch), and fragments for reuse.

In practice

• Use ACDL for documenting LLM agent prompt logic.
• Employ ACDL visualizations to compare context strategies.
• Utilize ACDL tooling for authoring and rendering specifications.

Topics

• Agentic LLM Contexts
• Context Description Language
• LLM Agent Systems
• Prompt Engineering
• Multi-turn Interactions

Best for: AI Architect, NLP Engineer, Research Scientist, AI Scientist, Machine Learning Engineer, AI Engineer

Related on AIssential

• 4 Things Everyone Must Know About LLMs and Agentic AI — Effective LLM and agentic AI development prioritizes model selection, task simplification, context management, and structural prompting over complex prompts.
• Agentic MPC for Semantic Control System Resynthesis — An agentic MPC framework integrates LLM-based agents to enable context-aware, semantically adaptive control synthesis.
• LLM-as-Code Agentic Programming for Agent Harness — Agentic Programming reassigns control flow from LLMs to the program, treating LLMs as "LLM-as-Code" for reliable, context-efficient reasoning.
• Article: From Prompts to Production: A Playbook for Agentic Development — Agentic AI development requires a specialized lifecycle and testing to manage nondeterministic behavior and ensure production readiness.
• Claude Code's Hidden Multi-Agent Orchestration now Open-source — AI agents are evolving to manage complex tasks and context more autonomously and efficiently.
• A Technical Taxonomy of LLM Agent Communication Protocols — A new taxonomy classifies LLM agent communication protocols, revealing architectural patterns and future trends.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.MA updates on arXiv.org.