Agent Orchestration Layers (AOL)
Context and Intent grow with each AOL layer. Context is what the agent knows. Intent is yours - your vision and goals, which the agent carries when context is sufficient.
What is Agent Orchestration Layer (AOL)?
Agent Orchestration Layer (AOL) is a framework that measures how deeply an AI agent is integrated into a project. It has 5 layers across 3 zones: Friction (AOL 1-2), Integration (AOL 3-4), and Autonomy (AOL 5). AOL measures two things: how much context your agent has about your project, and how well it carries your intent - your vision, goals, and standards. Both grow with each layer. At AOL 1, the agent has neither. At AOL 5, context is rich and your intent is fully implemented without being restated.
What are the three AOL zones?
Every project sits in one of three zones on the orchestration curve. The more structure you give the agent, the higher it climbs.
What are the five AOL layers?
Each layer builds on all the previous ones - AOL 5 contains everything from AOL 1 through 4. The full stack requires APL 3 or higher. At APL 1 and 2, the first layers are still achievable - through Custom Instructions, Project settings, and uploaded reference files in tools like Claude.ai or ChatGPT. The layer names below use Claude Code as the reference. Other tools apply the same concepts under different names.
No project files. No memory. The agent works from general knowledge only. Every session starts from zero.
Friction Zone+ CLAUDE.md. The agent knows where it is and what the basic rules are.
Friction Zone+ /docs + MEMORY.md. The agent knows your patterns and standards - and builds a project diary it maintains itself.
Integration Zone+ Skills. The agent has custom tools for repeatable tasks and executes them with precision.
Integration Zone+ Hooks. Automatic triggers. The agent checks its own work without being asked.
Autonomy ZoneWhy are they called layers, not levels?
Power and Access are scales - your position on them depends on your tools and permissions. AOL is different. You do not move to a higher AOL. You build layers into your project.
Once Power and Access are calibrated, adding layers is the practical next step. A low-power setup with high orchestration is a strong combination - the agent has full context and clear intent, but its actions remain limited and easy to review.
The AI Setup Snapshot measures your AOL alongside your APL and AAL, and shows you where the gaps are.
Take the AI Setup SnapshotThe AOL framework was developed by Michael Negele / Agent Builder Academy. All rights reserved.