Qdrant Memory

Qdrant Memory is an MCP server that provides a dual-layer knowledge graph implementation with semantic search capabilities powered by the Qdrant vector database. It connects language models in MCP clients to a structured graph storage backed by a local JSON file and vectorized records in Qdrant using OpenAI embeddings. Designed for developers building conversational agents, long-term memory assistants, and retrieval-augmented generation systems, the server allows AI clients to persist facts, entities, and relationships between conversation sessions. Whenever an entity or relation is created or modified, the server updates the local memory.json file and simultaneously computes OpenAI embeddings to update the Qdrant vector collection. Language models can then inspect the complete knowledge graph or query it through semantic similarity search to recall relevant contextual entities and observations based on meaning rather than exact keyword matches.

Category: AI Memory & Context

Tags: knowledge-graph, qdrant, rag, semantic search, vector-database

Visit Qdrant Memory

How to install and configure Qdrant Memory

1. Clone the repository and install dependencies by running npm install. 2. Build the project with npm run build. 3. Alternatively, build and run using Docker with docker build -t mcp-qdrant-memory .. 4. Add the server configuration to your MCP client settings under mcpServers: json { "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" } } } } 5. Replace the environment variable values with your OpenAI and Qdrant credentials, adjust the file paths, and restart your MCP client.

What you can do with Qdrant Memory

• Storing cross-session facts, user preferences, and project details by calling create_entities and add_observations. * Linking related concepts, projects, and people together by creating directed relationships between stored knowledge entities. * Performing similarity lookups with search_similar to retrieve contextually relevant memories matching vague or descriptive natural language prompts. * Inspecting the full memory architecture with read_graph to audit existing relationships and stored entity observations. * Deleting outdated observations, orphaned relations, or obsolete entities to keep conversational context accurate and synchronized.

Key facts

• Open Source
• https://github.com/delorenj/mcp-qdrant-memory
• AI Memory & Context, Databases & Data Stores
• knowledge-graph, qdrant, rag, semantic search, vector-database

Part of MCP Servers

What is Qdrant Memory?

Qdrant Memory is an open-source MCP server that provides knowledge graph storage combined with semantic search. It persists entities and relations to a local JSON file while storing corresponding OpenAI embeddings inside a Qdrant vector database.

What tools does Qdrant Memory provide?

It provides entity management tools including create_entities, create_relations, add_observations, delete_entities, delete_observations, delete_relations, and read_graph, along with the search_similar tool for vector-based semantic retrieval across entities and relations.

What environment variables are required to run Qdrant Memory?

You must provide OPENAI_API_KEY for embedding generation, QDRANT_URL to locate your Qdrant instance, QDRANT_COLLECTION_NAME for the target vector collection, and QDRANT_API_KEY if your Qdrant instance uses authentication.

What license does Qdrant Memory use?

Qdrant Memory is distributed under the open-source MIT license, allowing users to modify, deploy, and self-host the server for personal or commercial projects.