# Memory Box

> Memory Box is a universal memory layer for AI applications. It provides persistent, semantic memory for any LLM without relying on provider-native features. You own the memory, control the context, and hydrate any model on demand.

Key concepts:
- Provider-agnostic: Works with OpenAI, Anthropic, Google, or any LLM
- Semantic search: Vector, hybrid, and chronological retrieval modes
- Multi-tenant: Complete namespace isolation per agent
- Sub-50ms latency: Global edge infrastructure with Anycast routing
- Embedding: Qwen3-Embedding-8B with 1024 dimensions (included in API)

Base URL: https://memory-box-api.fly.dev

## API Reference

- [Authentication](https://memorybox.dev/docs#authentication): Two-layer auth with Bearer token and X-Agent-ID header
- [Base URL](https://memorybox.dev/docs#base-url): https://memory-box-api.fly.dev
- [Multi-Tenant Mode](https://memorybox.dev/docs#multi-tenant): Isolated tenants with prefixed API keys
- [Rate Limiting](https://memorybox.dev/docs#rate-limiting): 100 req/s for API, 10 req/min for Admin

## Memories

- [Store Memory](https://memorybox.dev/docs#store-memory): POST /api/v1/memories - Store and auto-embed memories
- [Search Memories](https://memorybox.dev/docs#search-memories): GET /api/v1/memories - Vector, hybrid, or chronological search
- [Get Memory](https://memorybox.dev/docs#get-memory): GET /api/v1/memories/{id} - Retrieve by ID
- [Delete Memory](https://memorybox.dev/docs#delete-memory): DELETE /api/v1/memories/{id} - Permanent deletion
- [Related Memories](https://memorybox.dev/docs#related-memories): GET /api/v1/memories/{id}/related - Semantic similarity

## Agents

- [Activate Agent](https://memorybox.dev/docs#activate-agent): POST /api/v1/agents/{agent_id}/activate - Create agent namespace
- [Agent Stats](https://memorybox.dev/docs#agent-stats): GET /api/v1/agents/{agent_id}/stats - Memory count, tokens, timing
- [Deactivate Agent](https://memorybox.dev/docs#deactivate-agent): DELETE /api/v1/agents/{agent_id} - Delete agent and all memories

## Admin (Multi-Tenant)

- [Create Tenant](https://memorybox.dev/docs#create-tenant): POST /admin/tenants - Create tenant with API key
- [List Tenants](https://memorybox.dev/docs#list-tenants): GET /admin/tenants - List all tenants
- [Get Tenant](https://memorybox.dev/docs#get-tenant): GET /admin/tenants/{id} - Get tenant details
- [Disable Tenant](https://memorybox.dev/docs#disable-tenant): DELETE /admin/tenants/{id} - Disable tenant access
- [Generate API Key](https://memorybox.dev/docs#generate-api-key): POST /admin/tenants/{id}/keys - Create new API key
- [List API Keys](https://memorybox.dev/docs#list-api-keys): GET /admin/tenants/{id}/keys - List tenant keys
- [Revoke API Key](https://memorybox.dev/docs#revoke-api-key): DELETE /admin/tenants/{id}/keys/{key_id} - Revoke key

## MCP Server

- [Overview](https://memorybox.dev/docs#mcp-overview): Claude Desktop and MCP client integration with OAuth 2.1
- [Discovery](https://memorybox.dev/docs#mcp-discovery): GET /mcp - Server information for client discovery
- [JSON-RPC](https://memorybox.dev/docs#mcp-json-rpc): POST /mcp - Main endpoint for MCP operations
- [Close Session](https://memorybox.dev/docs#mcp-close-session): DELETE /mcp - Close MCP session

## System

- [Health Check](https://memorybox.dev/docs#health): GET /health - Liveness check
- [Ready Check](https://memorybox.dev/docs#ready): GET /ready - Readiness with dependency status

## Architecture

- [Embedding Model](https://memorybox.dev/docs/architecture/embedding-model): Qwen3-Embedding-8B with 1024-dimensional vectors
- [Infrastructure & Scaling](https://memorybox.dev/docs/architecture/infrastructure-scaling): Fly.io global deployment with autoscaling

## Optional

- [FAQ](https://memorybox.dev/faq): Concepts, architecture, and how Memory Box differs from RAG
- [Security](https://memorybox.dev/security): SOC 2 Type II, GDPR compliance, data policies