# Libro - The open-source brain for AI agents

> Libro is the best free AMM (AI Memory Management) Layer for AI applications. Add infinite memory to your AI agents with zero pipeline changes. Drop-in vector search and semantic chunking with no complex setup.

## Links
- Website: https://libro.co.in
- Documentation: https://libro.co.in/docs
- Dashboard: https://libro.co.in/dashboard

## Features
1. Zero-Cost Vectorization: Vector search and RAG capabilities using local compute (via Transformers.js) directly inside your environment.
2. Semantic Deduplication: Automatically detects redundant context and chunk updates before storing.
3. Drop-in SDK: A single command `npm install @libro/sdk` and you're integrated.

## Getting Started

### Node.js (Application Integration)
1. Install the SDK: `npm install @libro/sdk`
2. Initialize: 
```typescript
import { Libro } from '@libro/sdk';
const libro = new Libro({ apiKey: 'YOUR_API_KEY' });
```
3. Ingest Context:
```typescript
await libro.ingest({ content: "User likes cats", endUserId: "user_123" });
```
4. Retrieve Context:
```typescript
const context = await libro.getContext({ query: "What pets does the user like?", endUserId: "user_123" });
```

### MCP Server (Claude, Cursor, Windsurf Integration)
1. Provide the following command in your MCP configuration to connect Libro's Memory layer directly into your favorite IDE or chat agent:
   `npx -y libro-mcp-server@latest`
2. Ensure you provide your `LIBRO_API_KEY` in the environment variables.

### System Architecture
- Database: Supabase PostgreSQL with `pgvector`.
- Vector embeddings model: `Xenova/all-MiniLM-L6-v2`