# Ryumem

> Semantic memory layer for AI agents - open source, self-hosted

Ryumem is an open-source semantic memory system that enables AI agents to store, search, and retrieve episodic memories with automatic knowledge graph extraction.

## Key Features

- Episodic memory storage with automatic entity/relationship extraction
- Hybrid search combining semantic, BM25, and graph traversal
- Multi-tenant isolation for user/session separation
- MCP server integration for Claude and other AI assistants
- Google ADK integration for Gemini-based agents

## Documentation

- [Introduction](/introduction): Overview of Ryumem and its capabilities
- [Quickstart](/quickstart): Get started in 5 minutes
- [Setup](/setup): Detailed installation and configuration

### Core Concepts

- [Architecture](/core-concepts/architecture): System design and components
- [Episodes](/core-concepts/episodes): Understanding episodic memory
- [Multi-tenancy](/core-concepts/multi-tenancy): User and session isolation
- [Search Strategies](/core-concepts/search-strategies): Semantic, BM25, hybrid, and traversal

### Integrations

- [MCP Server](/integrations/mcp-server): Integration with Claude Code and MCP clients
- [Google ADK](/integrations/google-adk): Integration with Google Agent Development Kit

### Examples

- [Basic Usage](/examples/basic-usage): Common usage patterns and code examples

## API Reference

The Ryumem MCP server exposes these tools:

- `add_episode`: Store new episodic memories
- `search_memory`: Multi-strategy semantic search
- `get_entity_context`: Retrieve entity and relationship data
- `list_episodes`: Paginated episode listing
- `get_episode`: Retrieve specific episode by UUID
- `update_episode_metadata`: Update episode metadata
- `prune_memories`: Clean up expired memories

## Source Code

GitHub: https://github.com/predictable-labs/ryumem

## Optional

For complete documentation content, see: /llms-full.txt