Justin0504

Justin0504 / Aegis

Public

Runtime policy enforcement for AI agents. Cryptographic audit trail, human-in-the-loop approvals, kill switch. Zero code changes.

ai-agents ai-safety anthropic audit-trail langchain
19
1
100% credibility
Found Mar 07, 2026 at 19 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

Aegis is a self-hosted safety system that intercepts AI agent actions before they happen, blocks risks, enables human approvals, and provides a real-time dashboard for monitoring.

How It Works

1
📰 Discover AI agent safety needs

You hear about powerful AI helpers that can accidentally delete files or send private data, and learn about Aegis to keep them safe.

2
🚀 Start Aegis easily

Download and launch Aegis on your computer with a single command — it sets up a safety guard for your AI projects.

3
🔗 Link your AI helper

Add one simple line to your AI code or set a quick option so every action goes through Aegis first.

4
🛡️ See safe actions in action

Run your AI helper and watch it check each step for danger before doing anything risky.

5
📊 Open the control dashboard

Visit the web page to see live updates on what your AI is doing, with charts and alerts.

6
Handle risky steps
✅
Allow safe ones

Click approve to let helpful actions continue.

❌
Stop dangers

Click block to prevent harmful mistakes.

🎉 Enjoy secure AI magic

Your AI helpers now work confidently with full oversight, alerts, and no surprises.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 19 to 19 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 Aegis?

Aegis adds runtime policy enforcement to AI agents, intercepting tool calls before execution to block risks like SQL injections, file traversals, or data exfiltration. It delivers cryptographic audit trails, human-in-the-loop approvals via a dashboard, and a global kill switch—all with zero code changes, just an env var or one import line. Built in TypeScript, it proxies calls across Anthropic, OpenAI, LangChain, and seven other frameworks via lightweight Python/JS/Go SDKs, spinning up via docker compose for a localhost dashboard and gateway.

Why is it gaining traction?

Unlike observability tools like LangFuse that log after damage, Aegis prevents issues pre-execution with auto-classification of any tool name and natural-language policy editing. Devs love the 30-second docker compose up -d setup, self-hosted SQLite backend, and features like OTLP export, Slack alerts, and Claude Desktop integration—far beyond aegis authenticator or aegis shield GitHub alternatives focused on auth or icons. The hook: tamper-proof hash-chained logs verify compliance without vendor data leaks.

Who should use this?

AI engineers building production agents with tools accessing databases, files, or networks—think LangGraph workflows querying prod DBs or CrewAI automating reports. Security teams enforcing agentcore runtime policy in .net runtime GitHub pipelines or arma 3 aegis GitHub mods needing kill switches. Anyone tired of post-mortem logs when deploying unpredictable agents.

Verdict

Grab it for agent safety prototyping—docker compose delivers instant value with polished docs and demo scripts, despite 19 stars and 1.0% credibility score signaling early maturity. Low test coverage means audit production deploys carefully, but it's a smart shield for rising AI risks.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.