Skip to content

Latest commit

 

History

History
90 lines (59 loc) · 4.07 KB

File metadata and controls

90 lines (59 loc) · 4.07 KB

Agent Workflows

Complete documentation of how AI agents interact with Marcus throughout their lifecycle.

Purpose

Understand the sophisticated multi-system orchestration behind every agent action. These guides reveal the intelligence, coordination, and learning happening when agents work with Marcus.

Audience

  • AI agents using Marcus for coordination
  • Developers building agent integrations
  • System architects understanding agent coordination
  • Anyone debugging agent behavior

Workflow Guides

Complete agent lifecycle from startup to continuous work loop. Understand the decision process, tool usage, and coordination patterns.

When to read: First document for understanding agent operations

How agents register with Marcus and become part of the coordination system. Covers profile creation, capability evaluation, and team integration.

Systems involved: Agent Management, Event System, AI Decision Engine, Memory System (5+ stages)

The sophisticated 8-stage process of task assignment. From project state refresh to AI-powered task selection to context building and instruction generation.

Systems involved: Agent Coordination, Project Management, AI Engine, Context System, Lease Management, Memory (15+ systems)

What happens when agents report progress at 25%, 50%, 75%, or 100%. Covers lease renewal, performance learning, predictive analytics, and cascade coordination.

Systems involved: Lease Management, Kanban Integration, Memory, Predictive Analytics, Monitoring (7+ stages)

AI-powered blocker analysis and resolution. Learn how Marcus analyzes root causes, generates solutions, assesses risk, and coordinates team response.

Systems involved: AI Blocker Analysis, Risk Assessment, Memory, Task Management, Communication Hub (7+ stages)

How agents retrieve comprehensive task context including dependencies, implementation patterns, architectural decisions, and risk assessment.

Systems involved: Core Models, Kanban Integration, Context System, Code Analysis, Memory, Risk Analysis (5+ stages)

Sophisticated dependency validation with graph analysis, status checking, predictive risk analysis, and optimization recommendations.

Systems involved: Context Analysis, Dependency Engine, Predictor Engine, Coordination Hub, Optimizer Engine, Learning (7+ stages)

What Makes These Guides Different

These aren't simple API docs—they reveal the internal complexity and intelligence behind each operation:

  • Multi-stage orchestration - 4-8 stages per operation
  • Multi-system integration - 8-15+ systems working together
  • AI-powered intelligence - Extensive AI analysis, prediction, optimization
  • 4-tier memory integration - Continuous learning at every step
  • Predictive analytics - Risk assessment, timeline forecasting, impact analysis
  • Comprehensive logging - Complete audit trail and observability

Agent Workflow Pattern

All agents follow this continuous loop:

1. Register (once) → 2. Request Task → 3. Get Context (if needed) →
4. Work on Task → 5. Report Progress (25%, 50%, 75%) →
6. Report Completion (100%) → 7. IMMEDIATELY Request Next Task → (loop to step 2)

Critical behaviors:

  • Complete tasks before requesting new ones
  • Request next task IMMEDIATELY after completion
  • Log decisions AS they're made
  • Report blockers with attempted solutions
  • Use context to understand dependencies

Next Steps


Remember: These operations look simple from outside, but they orchestrate sophisticated intelligence for effective coordination.