Type-safe clients for all four AIRS service domains — real-time content scanning, security configuration management, model security analysis, and AI red teaming. Zero external HTTP dependencies. Native fetch + crypto. ESM-first with dual CJS/ESM exports.
Features
- Real-Time Content Scanning — Synchronous and asynchronous scanning of AI prompts and responses. Detect prompt injection, toxic content, data leaks, and malicious URLs inline.
- Security Management — Full CRUD for security profiles and custom topics. Programmatically manage your AIRS configuration at scale.
- Model Security — Scan ML models for supply chain threats: malicious code execution, backdoors, unapproved file formats. Manage security groups and rules.
- AI Red Teaming — Run automated red team scans against AI targets. Static attack libraries, dynamic agent-based testing, custom prompt sets, and comprehensive reporting.
- Type-Safe Everything — 50 typed enum const objects, 250 Zod schemas with
.passthrough()for forward compatibility, and JSDoc on every exported symbol. - Zero Dependencies — Native
fetch+cryptoonly. No external HTTP libraries. Exponential backoff retry built in. ES2022 target, Node 18+.
Four Independent APIs
Get Started
- Install — Install the SDK, configure credentials, and connect to AIRS.
- Quick Start — Scanning, management, and more in under 5 minutes.
- Configure — Environment variables, auth methods, and endpoint setup.
- API Reference — Complete method signatures, types, enums, and error handling.