home / mcp / mcp memory server with qdrant persistence mcp server
MCP server providing a knowledge graph implementation with semantic search capabilities powered by Qdrant vector database
Configuration
View docs{
"mcpServers": {
"delorenj-mcp-qdrant-memory": {
"command": "/bin/zsh",
"args": [
"-c",
"cd /path/to/server && node dist/index.js"
],
"env": {
"QDRANT_URL": "http://your-qdrant-server:6333",
"OPENAI_API_KEY": "your-openai-api-key",
"QDRANT_API_KEY": "your-qdrant-api-key",
"QDRANT_COLLECTION_NAME": "your-collection-name"
}
}
}
}You run an MCP Memory Server with Qdrant persistence to store a knowledge graph in memory-backed storage while leveraging Qdrant for fast semantic search. It uses OpenAI embeddings to enable semantic similarity, supports HTTPS and reverse proxies, and can be deployed via Docker for easy hosting and scaling.
You interact with the server through an MCP client to manage entities, relations, observations, and to perform semantic searches. Core capabilities include creating and linking entities, adding observations to enrich concepts, deleting items, retrieving the full knowledge graph, and searching for semantically similar concepts or relations.
Prerequisites: ensure you have Node.js and npm installed, and Docker if you plan to run the Docker image. You should also prepare a Qdrant instance and an OpenAI API key for embeddings.
Local installation steps (build and run from source):
npm install
npm run buildConfiguration is handled by MCP server settings. The following stdio-based setup runs the server locally by invoking the built distribution from a script path. You will need to customize the home path to your server location.
{
"mcpServers": {
"memory": {
"command": "/bin/zsh",
"args": ["-c", "cd /path/to/server && node dist/index.js"],
"env": {
"OPENAI_API_KEY": "your-openai-api-key",
"QDRANT_API_KEY": "your-qdrant-api-key",
"QDRANT_URL": "http://your-qdrant-server:6333",
"QDRANT_COLLECTION_NAME": "your-collection-name"
},
"alwaysAllow": [
"create_entities",
"create_relations",
"add_observations",
"delete_entities",
"delete_observations",
"delete_relations",
"read_graph",
"search_similar"
]
}
}
}The server supports HTTPS and can work behind reverse proxies. Use proper SSL/TLS configurations and verify certificates when connecting to Qdrant or exposing the MCP port. If you are running Qdrant behind a reverse proxy, ensure the proxy forwards the correct host and client IP headers.
If you encounter connectivity or embedding issues, verify that your OpenAI API key and Qdrant URL are correct, and that network paths between the MCP server, Qdrant, and OpenAI are open.
The server maintains two persistence layers: a local file-based store (memory.json) and a Qdrant vector store for embeddings. Changes synchronize across both layers to keep data consistent.
Create multiple new entities in the knowledge graph.
Create relations between existing entities to form a connected graph.
Attach observations to entities to enrich their context.
Remove entities and their associated relations from the graph.
Delete specific observations from an entity.
Delete specific relationships between entities.
Retrieve the full knowledge graph with entities and relations.
Find semantically similar entities and relations using embeddings.