MCP Memory Graph Server

What is MCP Memory Graph Server?

The MCP Memory Graph Server provides persistent storage for knowledge graphs using MongoDB. It enables AI agents to store entities, manage relationships, query graph structures, and receive real-time updates.

How to use MCP Memory Graph Server?

Install uvx and set up a MongoDB connection string via the environment variable MCP_MONGO_MEMORY_CONNECTION. Add the server to your MCP client’s configuration with the command uvx mongo-memory and the required connection string.

Key features of MCP Memory Graph Server

• MongoDB-based persistent storage
• Graph operations (create, read, update, delete)
• Entity management with unique names
• Relationship handling between entities
• Query capabilities with filtering
• Real-time updates and modifications

Use cases of MCP Memory Graph Server

• Storing entities and their properties for a knowledge base
• Managing relationships between entities for graph traversal
• Querying and retrieving graph structures for AI agents
• Real-time updates to memory during agent interactions

FAQ from MCP Memory Graph Server

What is required to run the server?

You need uvx installed and a running MongoDB instance (local, cloud, or MongoDB Atlas).

How do I configure the connection to MongoDB?

Set the environment variable MCP_MONGO_MEMORY_CONNECTION to your MongoDB connection string (e.g., mongodb+srv://...).

What operations are available for AI agents?

The server provides entity management (create_entities, get_entity, update_entity, delete_entity, find_entities), relationship management (create_relationship, get_relationships, delete_relationship), and memory structure tools (get_memory_structure, get_usage_guide).

Where is the data stored?

All data is stored in the MongoDB database specified in the connection string.

Can I use a local MongoDB instance?

Yes, any valid MongoDB URI—local, remote, or Atlas—works as long as it is reachable from the server host.