Kai Agent

AI security agents for your entire codebase

Kai hardens AI-generated and rapidly written code by finding, proving, and patching real vulnerabilities before they reach production. Turn vibe-coded prototypes into production-grade security.

Features

🎯 Orchestrated Intelligence for Security

Kai orchestrates specialized agents across multiple frontier models, each optimized for discovery, verification, or patching. You get the right capability at the right cost for every vulnerability, automatically.

✅ Verified by Execution, Not Pattern Matching

Every exploit is proven by execution. Every fix is validated the same way. If Kai can't reproduce an issue or confirm a patch, it marks the result as unverified. It never claims a win it can't prove.

⚡ Parallel Agents Turn Days into Hours

A single Kai runtime launches dozens of agents in parallel, scanning code, generating exploits, and iterating on fixes. What used to take a team 36 hours of focused work now completes in about 2 to 3 hours end to end.

☁️ Cloud Runtime, IDE Optional

Kai runs in isolated cloud sandboxes. Kick off a scan from your IDE or CI and walk away. Progress, live metrics, exploit traces, and suggested code diffs are all available in the web dashboard.

📊 Watch Security Evolve in Real Time

Watch vulnerabilities move from "found" to "reproduced" to "patched" in real time. Use the timeline view to see how scaling inference changes the security posture of your codebase as Kai runs.

Getting Started

1. Install the Extension: Install Kai Agent from the VS Code Marketplace
2. Sign In: Open the Kai panel in the Activity Bar and sign in with your account
3. Select Files: Choose the files you want to scan in the "Files to Review" view
4. Run Scan: Click the scan button and select your scan mode:
   • Baseline: Quick scan with basic analysis
   • Enhanced: Balanced scan with comprehensive analysis (recommended)
   • Full: Deep scan with maximum thoroughness
5. Review Results: View detailed exploit traces and suggested patches in the Scan Results view

Configuration

Access settings via the Kai Settings button or VS Code's Settings UI:

• OpenRouter API Key (Optional): Provide your own API key for LLM inference to reduce charges. Otherwise, Kai uses its own keys.
• Default Scan Mode: Choose your preferred default scan mode (Baseline, Enhanced, or Full)

Requirements

• VS Code 1.80.0 or higher
• Active internet connection for cloud runtime
• Kai account (sign up during first use)

How It Works

1. Discovery: Specialized agents scan your codebase for potential vulnerabilities
2. Verification: Exploits are generated and executed to prove the vulnerability exists
3. Patching: Fixes are proposed and validated through execution
4. Reporting: Detailed results with exploit traces and code diffs are provided

Support

• Report issues: GitHub Issues
• Documentation: kai.dria.co

About

Built by Dria. Team includes alumni from Amazon, Microsoft, and Morgan Stanley.

License

MIT License - see LICENSE file for details

Turn your security scanning from days to hours with AI agents that prove what they find.