Home
Context-aware security analysis for Visual Studio Code
Caspian Security detects vulnerabilities, insecure coding patterns, and security best practice violations as you write code. It provides 164 security rules across 14 categories, covering SQL injection, XSS, hardcoded secrets, business logic flaws, and more.
What sets it apart: context-aware intelligence. The scanner classifies issues with confidence scores, AI fixes understand full function scope, teams share ignore decisions via .caspianignore, and results export to SARIF v2.1.0 for GitHub Security Alerts.
| Rules | 164 (74 code-detectable + 59 informational) |
| Categories | 14 security categories |
| Languages | JavaScript, TypeScript, Python, Java, C#, PHP, Go, Rust |
| AI Providers | Anthropic Claude, OpenAI GPT-4, Google Gemini |
| Export Formats | JSON, CSV, SARIF v2.1.0 |
| Page | Description |
|---|---|
| Getting Started | Installation, first scan, scan modes |
| Configuration | Settings, AI providers, category toggles |
| AI Fixes | Smart context AI fix generation |
| Confidence Scoring | Critical / Safe / Verify Needed classification |
| Caspianignore |
.caspianignore file format and team workflow |
| SARIF Export | SARIF v2.1.0 export and GitHub Security integration |
| Rule Reference | Complete list of all 164 rules |
| FAQ | Troubleshooting and common questions |
- Context-aware analysis -- classifies issues by variable source with confidence badges
- AI fixes with function-level understanding -- sends the entire enclosing function to the AI
- 164 security rules across 14 categories with actionable fix suggestions
- Real-time analysis -- checks code as you type with a 1-second debounce
- Full workspace scanning -- scans all project files, not just open tabs
- 8 languages supported -- JS, TS, Python, Java, C#, PHP, Go, Rust
-
Team-shareable
.caspianignore-- version-controlled ignore decisions - SARIF v2.1.0 export -- upload to GitHub Security Alerts
- Per-category toggles -- enable or disable each category independently
- 3 AI providers -- Claude, GPT-4, Gemini
- Cancellable scans -- workspace scans show progress and can be stopped
- Configurable severity -- filter by error, warning, or info threshold