AIProxyGuard

LLM Security Proxy with Prompt Injection Detection.

What is AIProxyGuard?

AIProxyGuard is a security proxy that sits between your application and LLM providers (OpenAI, Anthropic, etc.) to detect and block prompt injection attacks, jailbreak attempts, and other malicious inputs before they reach the model.

You can use the Cloud API directly with our SDKs - no infrastructure to deploy or manage.

AIProxyGuard Architecture

Quick Start

# Pull and run
docker run -d -p 8080:8080 ghcr.io/ainvirion/aiproxyguard:latest

# Verify it's running
curl http://localhost:8080/healthz

Then point your LLM client to http://localhost:8080/openai/v1 instead of https://api.openai.com/v1.

Alternative: Also available on Docker Hub as ainvirion/aiproxyguard

Features

Feature Status Description
Multi-Provider Routing Production OpenAI, Anthropic, OpenRouter, Ollama
Request Scanning Production Professional signatures with sensibility adjustment
Response Scanning Production Detect sensitive data leakage
Policy Engine Production Per-category actions with thresholds
Prometheus Metrics Production Full observability at /metrics
Health Endpoints Production /healthz, /readyz for orchestration
Control Plane Beta Fleet management, signature sync
TLS Interception Beta HTTPS inspection via MITM proxy

Detection Categories

Category Description Default
prompt-injection Instruction override attempts Block
jailbreak DAN mode, persona exploits Block
encoding-bypass Base64/hex/ROT13 obfuscation Block
delimiter-injection JSON/XML/markdown structure attacks Block
indirect-injection Tool abuse, plugin exploits Block
unicode-evasion Homoglyphs, fullwidth chars Block
role-manipulation Named character roleplay attacks Block

Next Steps