What is Azen?
Azen is an AI memory infrastructure that helps developers build stateful AI applications by providing:- Secure memory storage with AES-256-GCM encryption at rest
- Semantic search using vector embeddings for intelligent memory retrieval
- Organization-based multi-tenancy for team collaboration
- API key management with configurable rate limiting
- Usage tracking and analytics for monitoring and billing
Quickstart
Get up and running with Azen in under 5 minutes
API Reference
Explore the complete API documentation
Core Concepts
Learn how Azen’s memory system works
Authentication
Set up API authentication for your application
Key Features
Encrypted Memory Storage
Memory text is encrypted at rest in Postgres using AES-256-GCM. Plaintext exists only in-memory during requests.
Semantic Search
Find relevant memories using natural language queries powered by vector embeddings and similarity search.
Multi-Tenancy
Organization-based tenancy with role-based access control and team collaboration features.
Rate Limiting
Configurable rate limits per API key with automatic token bucket refills and usage tracking.
Async Processing
Memory embeddings are processed asynchronously using BullMQ for reliable, scalable operations.
Usage Analytics
Track API usage, request counts, and memory operations per organization for monitoring and billing.
Architecture
Azen is built as a secure-by-default memory infrastructure:- Memory encryption: All memory text is encrypted at rest in Postgres. Decryption happens only during active requests.
- Vector separation: Embeddings are stored separately in Pinecone, isolated from encrypted plaintext.
- Ephemeral processing: Workers operate on temporary payloads without persistent access to decrypted memory.
Security is treated as a baseline requirement, not an add-on feature.
Technology Stack
Azen is built with modern, production-ready technologies:- Backend: Hono (TypeScript API framework)
- Database: Postgres with Drizzle ORM
- Vector DB: Pinecone for semantic search
- Auth: Better Auth for user authentication
- Queue: BullMQ with Redis for async processing
- Runtime: Bun for fast TypeScript execution
Use Cases
Azen is designed for AI applications that need persistent, searchable user context:- AI assistants that remember user preferences and conversation history
- Personalized recommendations based on stored user behavior and interests
- Context-aware chatbots that retrieve relevant information from past interactions
- Knowledge management systems with semantic search capabilities
Getting Started
Create an account
Sign up at azen.sh and create your first organization.
Ready to start?
Follow our quickstart guide to create your first memory