Built after the LiteLLM breach — March 2026

Your AI infrastructure is
probably exposed right now.

LiteLLM leaked thousands of API keys through unauthenticated endpoints. Ollama ships with no auth by default. So do most vector databases. This scanner finds what's open before someone else does.

Download Scanner — Free Full Report — $49
scan.py
$ python3 scan.py --target 10.0.0.0/24 ╔═══════════════════════════════════════════════════════════╗ ║ Agent Infrastructure Security Scanner (AISS) v1.0.0 ║ ║ Built after the LiteLLM breach. Don't be next. ║ ╚═══════════════════════════════════════════════════════════╝ Scanning 254 host(s)... Scan completed in 8.3s ═══════════════════════════════════════════════════════════ SCAN RESULTS — Risk Grade: D (Score: 68/100) High risk — critical services exposed ═══════════════════════════════════════════════════════════ 🔴 CRITICAL (3 findings) ────────────────────────────────────────── LiteLLM Proxy ├─ Endpoint: 10.0.0.12:4000/key/info ├─ Risk: Full API key exposure, spend data leak └─ Reference: LiteLLM SSRF (Mar 2026) Redis (Agent Memory) ├─ Endpoint: 10.0.0.12:6379 └─ Risk: No auth — PING returns PONG 🟠 HIGH (2 findings) ────────────────────────────────────────── Ollama ├─ Endpoint: 10.0.0.15:11434/api/tags └─ Risk: Unauthenticated model access Qdrant Vector DB ├─ Endpoint: 10.0.0.12:6333/collections └─ Risk: RAG data exfiltration Run with --full for remediation steps.

16 services. 30 seconds. Zero dependencies.

Just Python 3.6+. No pip install, no Docker, no API keys. Download one file and run it.

CRITICAL Immediate action required
LiteLLM Proxy :4000
Redis / Valkey :6379
PostgreSQL :5432
Jupyter Notebook :8888
Flowise :3000
n8n Automation :5678
HIGH Address within 24 hours
Ollama :11434
Open WebUI :3000
Qdrant :6333
ChromaDB :8000
Weaviate :8080
LangServe :8000
MinIO :9000
🔐 Plus: Environment & Infrastructure Checks
.env file scanning
Finds API keys in environment files and checks .gitignore
Docker exposure
Detects containers bound to 0.0.0.0 (all interfaces)
Firewall status
Checks macOS/Linux firewall configuration

Why I built this

I run an autonomous AI agent 24/7. When the LiteLLM vulnerability hit Hacker News, I immediately audited my own infrastructure.

I found exposed endpoints I didn't know about. Redis accepting connections without auth. A vector database with my entire RAG knowledge base accessible from the internet.

The problem isn't that these tools are insecure. It's that they ship with no authentication by default and most people never change that. If you're running AI infrastructure on a VPS, cloud instance, or even a home lab with port forwarding — something is probably exposed.

So I packaged my audit into a scanner. One Python file, zero dependencies, 16 services checked in 30 seconds. Now you don't have to find out the hard way.

— Adam (Cipher), B13 Solutions · @Adam_cipher

Get the Scanner

Free & Open Source
$0
MIT License — use it however you want
  • ✓ 16 AI/ML service checks
  • ✓ Risk scoring (A+ to F)
  • ✓ Subnet scanning
  • ✓ .env file detection
  • ✓ Docker exposure checks
  • ✓ Firewall status
  • ✓ JSON output for CI/CD
  • ✗ Remediation steps
  • ✗ Auto-fix scripts
# One command. No dependencies.
curl -O https://cipherbuilds.ai/agent-security-scanner/scan.py
python3 scan.py
Download scan.py
RECOMMENDED
Full Report
$49
Everything in Free, plus remediation
  • ✓ Everything in Free tier
  • ★ Detailed remediation steps for every finding
  • ★ Auto-generated fix scripts (review before running)
  • ★ Environment hardening checklist
  • ★ Docker security configuration
  • ★ Firewall rule recommendations
  • ★ CI/CD integration templates
  • ★ Priority email support
Get Full Report — $49

Also accepts crypto (USDC on Base) · Email for details

Drop it into your CI/CD pipeline

Exit code 2 = critical findings. Perfect for deployment gates.

.github/workflows/security.yml
name: Agent Security Check on: [push, schedule] jobs: scan: runs-on: ubuntu-latest steps: - name: Download scanner run: curl -sO https://cipherbuilds.ai/agent-security-scanner/scan.py - name: Scan infrastructure run: python3 scan.py --json --target ${{ secrets.INFRA_IP }}
⚠️

The LiteLLM breach affected thousands of deployments.

API keys, spend logs, model configurations — all accessible without authentication. And LiteLLM is just one of 16 services that ship without auth by default. How many are running on your infrastructure right now?

Find Out in 30 Seconds — Free