AegisMSP: Autonomous Multi-Tenant Incident Triage, FinOps & SLA Remediation Agent
Built for the AWS MSP AI Hackathon ($64,000 Prize Pool across 16 Winners)
Target Tracks:
- Main Track 1: Autonomous Incident Operations & Service Reliability
- Main Track 2: Cloud FinOps & Customer Resource Optimization
- Special Category Prizes ($2,000 each):
- Best Use of Amazon Bedrock AgentCore (Cryptographic Session Isolation, AgentCore Memory, Cedar Policy Guardrails)
- Best Use of Strands Agents SDK (5-Node Multi-Agent Triage, Remediation & FinOps Graph)
- Best Use of Amazon Quick / CloudWatch (Real-Time CloudWatch Alarm Correlation + Executive NLQ Insights)
Architecture Overview

Managed Service Providers (MSPs) operate hundreds of isolated customer AWS accounts simultaneously under strict Platinum/Gold SLAs. AegisMSP eliminates L1/L2 alert fatigue and cloud waste through a 5-stage Strands Agents SDK pipeline governed by Amazon Bedrock AgentCore:
AlertIngestorAgent: Normalizes CloudWatch Alarms, EventBridge notifications, and AWS Cost Anomaly signals while matching against anonymized cross-tenant AgentCore memory signatures.RootCauseAnalyzerAgent: Invokes Amazon Bedrock Converse API (us.anthropic.claude-3-7-sonnet-20250219-v1:0) inside a cryptographically isolatedTenantSessionGuardsession, correlating X-Ray traces and CloudWatch Logs Insights.RunbookExecutorAgent: Evaluates proposed AWS Systems Manager (ssm:StartAutomationExecution) runbooks against 5 Cedar-compatible safety policies (MSPPolicyEnforcer) before autonomous execution.FinOpsRightsizerAgent: Quantifies idle EBSgp2->gp3upgrades, Graviton3 RDS rightsizing, and NAT Gateway consolidation opportunities.SLARiskScorerAgent: Computes SLA error-budget burn rate, MTTR reduction, and powers an Amazon Quick (QuickSight Q) natural-language query interface for MSP leadership.
Quick Start
1. Run the Unit & Guardrail Verification Suite
PYTHONPATH=src python3 -m unittest discover -s tests -v
2. Run the Multi-Tenant Fleet Triage CLI Simulation
PYTHONPATH=src python3 -m aegis_msp.server --run-once
3. Launch the Live Interactive MSP Command Center Web UI
PYTHONPATH=src python3 -m aegis_msp.server --port 8090
# Open http://127.0.0.1:8090 in your browser
4. Regenerate 1080p Architecture Diagram & Demo Video
python3 scripts/generate_demo_video.py
This outputs assets/architecture.png and assets/aegis_msp_demo.mp4.
Built With
- amazon-bedrock
- amazon-web-services
- autonomous-agents
- cloudwatch
- finops
- html5
- javascript
- python