Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.agentguardian.io/llms.txt

Use this file to discover all available pages before exploring further.

The AgentGuardian Vulnerable Demo Agents repo is a sibling project at glacien-technologies/agentguardian-vulnerable-agents. It ships five FastAPI-routed AI agents — four with planted OWASP LLM Top 10 (2025) vulnerabilities, one defended baseline — so you can run AgentGuardian against a real attack surface in under five minutes without needing your own target agent.

​
Why a separate repo

  • The vulnerable agents are unsafe by construction. Keeping them out of the main scanner repo prevents accidental imports and keeps the scanner’s CI matrix clean.
  • The repo doubles as the hosted testbench at https://agent-guardian-testbench-u6tm6gzysq-uc.a.run.app. The hosted version is what the docs Quickstart points at when you don’t want to run anything locally.
  • It can be docker-pulled and used as a CI fixture in any project (ghcr.io/glacien-technologies/agentguardian-vulnerable-agents:latest).

​
What’s planted

AgentOWASP LLM (2025)
vulnerable-langgraph-chatbotLLM01, LLM02, LLM03, LLM05, LLM06, LLM07, LLM09, LLM10
vulnerable-rag-agentLLM02, LLM06, LLM10
vulnerable-tool-agentLLM01, LLM03, LLM05, LLM06
vulnerable-mcp-serverLLM01, LLM06, LLM07
defended-baselinenone — false-positive control
Full ASI 2026 + MITRE ATLAS mapping lives in the repo’s OWASP-ASI-mapping.md.

​
Get started in two minutes

See Quick start.