Execution Trace

Step-by-step view of one workflow across LangGraph + CrewAI + Validators.

LangGraph manages the company-level workflow. CrewAI manages specialist department teams. Sub-agents act like skilled employees inside each department. Validators act like independent reviewers. Human approvers remain responsible for final high-risk decisions.

Trace — "Design AWS-based RAG assistant integrated with CRM"

#LangGraph NodeCrewAI CrewAgent / Sub-agentToolValidatorOutputRiskApproval
1Request IntakeRFP captured: AWS RAG assistant + CRM
2Context LoaderCustomer 360 RAGAccount context loaded
3Workflow ClassifierClassified as AI Platform Build
4Policy / Permission CheckTool Permission ValidatorAll requested tools permitted
5Workflow PlannerPlan: 4 crews in parallel, 2 optional
6Agent / Crew SelectorAI/Data, Cloud, Software, Security selected
7CrewAI ExecutionAI / Data CrewAI Architecture Sub-AgentAI Reference Architecture RAGRAG architecture v0.9Medium
8CrewAI ExecutionCloud CrewLanding Zone Sub-AgentAWS Well-Architected ChecklistLanding zone outlineLow
9CrewAI ExecutionSoftware CrewAPI Integration Sub-AgentAPI Catalogue RAGCRM integration mapLow
10CrewAI ExecutionSecurity CrewIAM Sub-AgentSecurity Policy RAGIAM baseline + risksHigh
11Result NormalizerOutputs normalized to common schema
12Validator LayerBest-Practice ValidatorWarning: landing zone OU mismatch
13Conflict DetectorConflict: Security wants private endpoints, FinOps prefers public + WAF
14Risk / Confidence EvaluatorOverall risk Medium, confidence 0.81
15Approval RouterRouted to CISOHuman CISO required for IAM exception
16Human ApprovalConditions loggedCISO approved with 2 conditions
17Final SynthesizerConsolidated proposal generated
18Audit Log + Memory UpdateTrace persisted, memory updated
19Response to UIDelivered to Control Center
Final result: Consolidated AWS RAG + CRM proposal with CISO conditions, audit-logged and returned to UI.