Adrian

Open-source runtime AI agent security tool - monitors and controls AI agents, catching malicious tool use, prompt injection, and policy drift in real time, befo

What is Adrian?

Open-source runtime AI agent security tool - monitors and controls AI agents, catching malicious tool use, prompt injection, and policy drift in real time, before the agent acts.

Overview of Adrian

What it is

Adrian is an open-source, AARM-aligned runtime security monitoring and control engine for AI agents. It analyses both agent activity logs (tool calls, actions, outputs) and reasoning traces to detect malicious, misaligned, or out-of-remit behaviour, and optionally intervene in-flight. It is described as an open-source runtime AI agent security tool that monitors and controls AI agents, catching malicious tool use, prompt injection, and policy drift in real time, before the agent acts.

Key features

  • Analyses both agent activity logs and reasoning traces to detect attacks.
  • Detects prompt injection, jailbreaks, tool poisoning, data exfiltration, privilege escalation, and out-of-remit actions.
  • Optionally intervenes in-flight by blocking or holding actions for approval.
  • Provides SDKs for Python (LangChain, Anthropic) and TypeScript.
  • Includes a native Claude Code plugin that secures every tool call from the terminal.
  • Uses classifiers that understand the agent's intended behavior, not just known attack patterns.
  • According to OpenAI and DeepMind research, achieves +35% detection accuracy and catches 4x more nuanced attacks compared to behaviour-only monitoring.

Use cases

  • Catching prompt injection and jailbreaks (direct and indirect).
  • Detecting tool poisoning and unsafe or off-policy tool calls.
  • Preventing data exfiltration and secret/credential leakage.
  • Identifying privilege escalation and out-of-remit actions.
  • Securing AI agents in production environments at runtime.

Getting started

sh pip install adrian-sdk pip install langchain langchain-openai # or langchain-anthropic, etc. For self-hosting with Docker: sh git clone https://github.com/secureagentics/Adrian cd Adrian docker compose --profile setup run --rm setup bootstrap docker compose --profile llm up -d

License and pricing

Licensed under Apache-2.0. Open source and free forever.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/secureagenticsadrian.svg)](https://yunodb.com/tools/secureagenticsadrian?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/secureagenticsadrian?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/secureagenticsadrian.svg" alt="Featured on YunoDB" height="20"></a>