Break complex tasks into sequential LLM calls
RoutingReflectionGuardrails & Safety
Direct inputs to specialized handlers
Prompt ChainingMulti-Agent CollaborationGuardrails & Safety
Execute independent tasks concurrently
PlanningMulti-Agent CollaborationEvaluation & Monitoring
Decompose goals into actionable steps before execution
ReflectionTool Use / Function CallingMemory Management
Enable agents to interact with external systems
PlanningGuardrails & SafetyException Handling & Recovery
Ground responses in external knowledge
Tool Use / Function CallingMemory ManagementGuardrails & Safety
Maintain context across interactions
ReflectionLearning & AdaptationKnowledge Retrieval (RAG)
Self-evaluate and iteratively improve outputs
PlanningPrompt ChainingEvaluation & Monitoring
Evolve behavior based on feedback
Memory ManagementReflectionEvaluation & Monitoring
Specialized agents working together
RoutingPlanningParallelization
Standardized tool and resource integration
Tool Use / Function CallingMulti-Agent CollaborationAgent-to-Agent Protocol (A2A)
Standardized inter-agent communication
Multi-Agent CollaborationModel Context Protocol (MCP)Routing
Graceful degradation and error recovery
Tool Use / Function CallingPlanningGuardrails & Safety
Strategic human oversight at critical points
Guardrails & SafetyTool Use / Function CallingPlanning
Prevent harmful or unwanted behaviors
Human-in-the-Loop (HITL)RoutingPrompt Chaining
Track progress toward objectives
PlanningHuman-in-the-Loop (HITL)Evaluation & Monitoring
Optimize for cost, latency, and quality
RoutingPlanningEvaluation & Monitoring
Enhance reasoning with structured techniques
ReflectionPlanningEvaluation & Monitoring
Measure and improve agent performance
ReflectionLearning & AdaptationHuman-in-the-Loop (HITL)
Rank and select tasks intelligently
PlanningGoal Setting & MonitoringResource-Aware Optimization
Autonomously seek novel information
PlanningTool Use / Function CallingReflection
