Concepts#

This section explains what Marcus is and why it works the way it does. These documents provide high-level understanding without implementation details.

Purpose#

Learn Marcus’s design philosophy, core principles, and fundamental concepts. Perfect for understanding the “big picture” before diving into how-to guides or technical systems.

Audience#

  • New users wanting to understand Marcus

  • Developers evaluating Marcus for their projects

  • Anyone interested in multi-agent coordination philosophy

  • Researchers studying AI agent systems

Documents#

Philosophy#

Marcus’s Stoic approach to multi-agent software development. Learn about BYOA (Bring Your Own Agent), context over control, and why randomness is a feature.

Core Values#

The seven principles that guide Marcus’s design: Sacred Repository, Guided Autonomy, Embrace Emergence, Relentless Focus, Radical Transparency, Context Compounds, and Fail Forward.

Hierarchical Task Decomposition#

How Marcus intelligently breaks down large, complex tasks into manageable subtasks with clear interfaces, dependencies, and shared conventions for effective agent collaboration.

Contract-First Decomposition#

How Marcus uses interface contracts to solve the Single-Author Problem in tightly-coupled multi-agent projects. Covers the motivation, experimental validation, architectural decisions, and what remains to be built.

Activity Tracking vs. Diagnostics#

The philosophy behind Marcus’s separation of activity tracking (recording what happened) and diagnostics (analyzing why it happened). Learn why mixing these concerns leads to false assumptions and how proper separation improves maintainability.

Key Concepts#

For detailed concept explanations, see the Getting Started - Core Concepts guide, which covers:

  • Agents - AI workers and their lifecycle

  • Tasks - Work units, dependencies, and intelligence

  • Projects - Structured collections with phases

  • Kanban Boards - Visual project management

  • Context System - Rich task understanding

  • Dependencies - Task relationships and ordering

  • Memory & Learning - Four-tier learning architecture

  • AI Intelligence Engine - Hybrid decision-making

  • Hierarchical Task Decomposition - Breaking complex tasks into subtasks

Next Steps#

Remember: Marcus provides context, structure, and intelligence—then trusts agents to deliver. The magic happens in the space between structure and freedom.