Autonomous AI-powered log analysis and self-healing pipeline for AWS infrastructure.
Detects anomalies. Diagnoses root causes. Remediates — without waking anyone up.
SentinelOps-AI is a production-grade autonomous operations agent built on AWS. It continuously ingests logs from your infrastructure, uses an AI agent to understand what's going wrong, and triggers self-healing actions — all without human intervention.
No more 3 AM pages for issues that fix themselves. No more sifting through thousands of log lines. SentinelOps-AI watches, thinks, and acts.
AWS Infrastructure Logs
│
▼
┌───────────────────┐
│ Log Ingestion │ ← CloudWatch, S3, Kinesis
│ (app/) │
└────────┬──────────┘
│
▼
┌───────────────────┐
│ AI Agent Core │ ← Pattern detection, anomaly analysis,
│ (agent/) │ root cause reasoning via LLM
└────────┬──────────┘
│
▼
┌───────────────────┐
│ MCP Servers │ ← Tool layer: restarts, scaling,
│ (mcp_servers/) │ alerts, rollbacks, SNS/SQS triggers
└────────┬──────────┘
│
▼
┌───────────────────┐
│ Self-Healing │ ← Automated remediation executed
│ Actions │ against live AWS environment
└───────────────────┘
The loop:
- Logs stream in from AWS (CloudWatch, S3, Kinesis Data Streams)
- The AI agent parses and analyzes patterns — errors, latency spikes, service crashes
- It reasons about root cause using an LLM with context from your infra
- MCP servers expose safe, scoped tools the agent can invoke (restart service, scale ASG, trigger SNS alert, open incident)
- Remediation fires automatically — the agent closes the loop
SentinelOps-AI/
├── agent/ # Core AI agent — reasoning, decision engine, tool orchestration
├── app/ # Log ingestion layer + orchestration entrypoint
├── infra/ # AWS infrastructure as code (IaC)
├── mcp_servers/ # MCP tool servers — expose AWS actions to the agent
├── tests/ # Test suite
├── requirements.txt
└── README.md
| Layer | Technology |
|---|---|
| Cloud | AWS (CloudWatch, S3, Lambda, SNS, SQS, EC2/ECS) |
| AI Agent | Python + LLM (Claude / OpenAI) |
| Tool Protocol | Model Context Protocol (MCP) |
| Infrastructure | AWS CDK / Terraform |
| Language | Python 3.11+ |
- Zero-touch remediation — agent detects and fixes without human escalation
- MCP-based tool layer — clean, auditable interface between the AI and your AWS environment
- Root cause reasoning — not just pattern matching; the agent explains why it's taking action
- Pluggable actions — restart, scale, rollback, alert, or open an incident ticket
- Fully async pipeline — non-blocking log ingestion handles high-throughput environments
- Audit trail — every agent decision and action is logged for post-incident review
- Python 3.11+
- AWS account with appropriate IAM permissions
- API key for your chosen LLM provider
git clone https://github.com/vijayrajeshr/SentinelOps-AI.git
cd SentinelOps-AI
pip install -r requirements.txtaws configure
# Set your region, access key, and secret keypython app/main.pyThe agent will begin ingesting logs and monitoring your infrastructure immediately.
┌─────────────────────────────────────────────────────────┐
│ AWS Environment │
│ │
│ EC2 / ECS / Lambda ──► CloudWatch Logs │
│ │ │
│ Kinesis / S3 │
└───────────────────────────────┼─────────────────────────┘
│
┌───────────▼──────────┐
│ SentinelOps-AI │
│ │
│ ┌────────────────┐ │
│ │ Log Ingestion │ │
│ └───────┬────────┘ │
│ │ │
│ ┌───────▼────────┐ │
│ │ AI Agent │ │
│ │ (LLM Core) │ │
│ └───────┬────────┘ │
│ │ │
│ ┌───────▼────────┐ │
│ │ MCP Servers │ │
│ └───────┬────────┘ │
└──────────┼───────────┘
│
┌────────────────┼─────────────────┐
│ │ │
Restart Scale ASG SNS Alert /
Service Policies PagerDuty
pytest tests/- Slack / Teams integration for agent action summaries
- Multi-region support
- Grafana dashboard for agent activity
- Fine-tuned model for domain-specific log patterns
- Support for GCP and Azure log sources
MIT License — see LICENSE for details.
Vijay Rajesh R
GitHub · LinkedIn
"Don't page the engineer. Let the system fix itself."