secureagentics

secureagentics / Adrian

Public

Open-source runtime security monitoring and control for AI agents.

18
7
100% credibility
Found May 12, 2026 at 18 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Adrian is an open-source security monitor for AI agents that analyzes their actions and thoughts to detect and control risky behavior, with a simple dashboard and easy integration.

How It Works

1
🔍 Discover Adrian

You hear about Adrian, a helpful tool that watches your AI helpers to keep them safe and on track.

2
🚀 Get started quickly

Sign up at the online dashboard in a minute or set up on your own computer with a simple launch.

3
📦 Add safety to your AI agent

Install the easy helper kit and add just two lines to connect it to your AI agent's brain.

4
⚡ Run your agent

Start your AI agent as usual and watch live events and safety checks appear in the dashboard.

5
🛡️ Customize your safety rules

Set what your agent should do, get alerts for worries, and choose when to pause or block actions.

✅ Your AI agents are secure

Now you can use your smart helpers confidently, knowing they're monitored and protected from going off track.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 18 to 18 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is Adrian?

Adrian delivers runtime security for AI agents, scanning tool calls, actions, outputs, and reasoning traces to flag malicious or off-remit behavior, with options to alert, review, or block in real-time. Developers pip-install a Python SDK that hooks into LangChain or LangGraph agents with two lines of code, feeding events to a Go backend and Next.js dashboard for verdicts, timelines, and human-in-the-loop reviews. Self-host the full stack via Docker Compose on an NVIDIA GPU for offline Gemma models, or use the managed app.adrian.secureagentics.ai.

Why is it gaining traction?

It stands out by combining behavior logs with reasoning analysis—OpenAI/DeepMind research shows 35% better detection than logs alone—using contextual world models to catch novel risks like an e-commerce agent resetting passwords, beyond dataset-trained classifiers. The two-line SDK auto-instruments frameworks, Discord/Slack alerts fire instantly, and full self-hosting ensures data sovereignty as a github alternative open source self hosted option among github open source tools.

Who should use this?

LangChain/LangGraph builders deploying agents for e-commerce, finance, or customer support, where unchecked tool calls could leak data or execute harm. Security teams auditing production AI workflows, especially those preferring self-hosted github open source tools over cloud dependencies.

Verdict

Grab it for LangGraph agents needing instant security—SDK and Docker setup shine—but with 18 stars and 1.0% credibility score, it's early-stage; run your own red-team tests before prod. Solid docs and tests make it a low-risk experiment.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.