Skip to main content

About the project

With AI, the amount of code being written is at an all-time high. This is great! Solo builders are shipping products that used to need a whole team, and experiments that once took weeks now take hours. There’s so much more opportunity to build, learn, and launch.

But when apps ship faster, the attack surface keeps growing, and so do the vulnerabilities. We need a faster way to find and fix security issues. That’s what the Esper threat-modeler does.

Point Esper at your codebase and it will:

  1. Map your app: it draws a diagram of the parts of your app, how data moves between them, and where the trust boundaries are.
  2. Find threats: it checks each part using STRIDE, plus extra checks for AI and agent risks.
  3. Rank them: it scores each issue from Critical to Low so you know what to fix first.
  4. Show the fix: it points to the exact file and line and suggests the code change.

It works with 11 languages. It’s built as a skill for Hermes Agent and runs on GLM-5.3, an open-weight model, through OpenRouter. I chose GLM-5.3 because it has fewer security restrictions than Claude or OpenAI models, so it can walk through real attack paths in full.

Examples

Fun fact: Esper is named after the Esper computer in Blade Runner, the machine Deckard uses to zoom into a photo and spot details no one else can see.

Built with

openrouter

Prize Categories Entered

  • DigitalOcean

    Best Open-Source AI Project

    Submitted
  • Event prize

    Most on Theme

    Accepted