Stryct.ai Docs

Stryct runs security evaluations of AI agents in your CI. Add a Stryct Studio API key and the Stryct GitHub Action (coming soon) to your repo. The action finds your agents, models, tools, MCP servers and auth, generates attacks for them, runs them in a sandbox, and suggests fixes. It tests your governance policies too.

These docs cover the attacks we run, how we run them, and what the numbers mean. We also publish results from public runs on the leaderboard.

How we run our evaluations

Setup

Get started

Create an account, get an API key, add the GitHub Action.

Open page →
71 techniques

Attacks

Every attack we run, and how each one works.

Open page →
How we test

Methodology

How we run our security evaluations.

Open page →
What the numbers mean

Metrics

What the grade, ASR and FIR actually mean.

Open page →

Search Stryct.ai Docs

Loading docs…

    ↑↓ to navigate ↵ to open Esc to close