Aegisora v4.0 is live: Open-source runtime security for AI agents

Hey everyone, Eray here. 👋

Building AI agents is fun, but putting them in production is terrifying. Most tools just observe what happened after the agent messed up. I needed a way to control what an agent is allowed to do right before it happens.

So, I built Aegisora — an open-source runtime security and execution control layer for AI agents. (Think of it as an action-firewall: Allow, Block, or Escalate to a human).

Today, I'm super excited to announce a massive milestone: Aegisora 4.0 "Continuity Seal".

As a solo founder, I've been grinding to make this architecture rock-solid. This release is entirely focused on extreme stability, scalability, and uninterrupted performance so you can actually trust your AI infrastructure in production.

What’s under the hood?
🛡️ Intercepts tool/API calls before they execute.
⚖️ Evaluates based on risk, context, and custom policies.
⛓️ Append-only, hash-chained evidence log for every single decision.
💻 100% Open Source (MIT licensed).

I’d love for fellow builders to tear it apart, test it out, and give me some brutal feedback.

🔗 GitHub Release: https://lnkd.in/ejzTNPV2
🌐 Website: aegisora.org

If you are building with AI agents, how are you currently handling runtime security? Let's chat in the comments! 👇