AIOStack

Find Shadow AI in your cloud

AIOStack automatically discovers AI components you didn't know existed and ties each to an owner. No code changes required. Runs in-cluster via eBPF.

Features

Discovery

• Shadow AI across namespaces/clusters
• AI agents, MCP servers/clients, LLM endpoints, Vector DBs
• Orphaned/unused (“zombie”) AI services

Monitoring & Analytics

• Model/API calls (OpenAI, Anthropic, Gemini, etc.)
• Model downloads and runtime inventory
• Cost & usage insights

Security & Compliance

• Sensitive-data exposure cues and risky egress paths
• Evidence for investigations with owner attribution (Pods/Deployments/ServiceAccounts)

Supported Environments

• Kubernetes 1.29+ with eBPF support
  (EKS, GKE, AKS already satisfy this)
• Linux kernel 5.15+

Installation Requirements

• Helm 3.x
• kubectl configured

Quick Start

1. Create your Free Account

1. Sign up at app.aurva.ai (takes 30 seconds)
2. Copy your credentials from the email sent to you:
   • Company ID
   • AIOStack Validation Key

2) Install via Helm

Step 1: Configure Your Credentials

helm repo add aiostack https://charts.aurva.ai/
helm repo update

# Extract the default values file
helm show values aiostack/aiostack > values.yaml

Edit values.yaml and set your credentials/placeholders:

outpost:
  env:
    - name: COMPANY_ID
      value: "<YOUR_COMPANY_ID>"
    - name: AIOSTACK_VALIDATION_KEY
      value: "<YOUR_VALIDATION_KEY>"

observer:
  env:
    - name: IS_OUTPOST_URL_SECURE
      value: "false"

Optional: pin to the newest components by setting version: latest :

observer:
  version: latest
  ...

outpost:
  version: latest
  ...

Step 2: Deploy to Your Cluster

# Create namespace
kubectl create namespace aiostack

# Install with your configured values
helm install myaiostack aiostack/aiostack --namespace aiostack --values values.yaml

Step 3: Verify Installation

# Check if pods are running
kubectl get pods -n aiostack

Step 4: View the Dashboard

That's it! You can now access your Shadow AI inventory at app.aurva.io and login with your credentials (your username is the email you signed up with).

Supported Platforms

Platform	Status
EKS (AWS)	✅ Full
GKE (Google)	✅ Full
Kind/Minikube	✅ Dev only

Architecture

┌─────────────────┐    ┌──────────────────┐    ┌─────────────────┐
│   Applications  │───▶│   eBPF Agents    │───▶│  Data Pipeline  │
│                 │    │                  │    │                 │
│ • Python ML     │    │ • Syscall hooks  │    │ • Classification│
│ • Node.js AI    │    │ • Network trace  │    │ • Enrichment    │
│ • Java services │    │ • Process trace  │    │ • Aggregation   │
│ • Go binaries   │    │ • File I/O trace │    │                 │
└─────────────────┘    └──────────────────┘    └─────────────────┘
                                                        │
┌─────────────────┐    ┌──────────────────┐             │
│   Dashboards    │◀───│   ClickHouse     │◀────────────┘
│                 │    │                  │
│ • AIOStack UI   │    │ • Time-series DB │
│ • Web Console   │    │ • 7-30 day       │
│ • REST APIs     │    │   retention      │
└─────────────────┘    └──────────────────┘

Documentation & Support

• 📖 Documentation: https://aurva.ai/docs/home
• ⚙️ Detailed Installation Guide - https://aurva.ai/docs/installation/steps
• 🐛 Bug Reports: GitHub Issues
• 📧 Support: support@aurva.io

🔒 Security

Security Model

• eBPF programs use only required, minimal capabilities.
• Aligned with Kubernetes Security Contexts and Pod Security Standards.

Reporting Vulnerabilities: support@aurva.io
Security Audit: Results will be continously published (as available).

Data Handling and Privacy

• No TLS key access required; observability happens at syscall level.
• In-cluster operation. Data remains in your environment.
• Metadata Only:
  • Request/response bodies are not stored.
  • Sensitive values are classified in runtime.

License

AIOstack is open source.
Apache License 2.0 - see LICENSE for details.

⭐ Star History

---

Made with ❤️ by Aurva "See what others can't. Secure what others miss."