MCP Memory Server

MCP Memory Server is an MCP server that integrates Claude Desktop with Azure Cosmos DB and OpenAI embeddings to provide a persistent memory and context retrieval layer across multiple sessions. Designed for developers, researchers, and frequent AI users who require conversation continuity, the server monitors discussions, extracts relevant entities like people, projects, and technologies, and generates vector embeddings to index contextual data. It connects directly to an Azure Cosmos DB database containing dedicated containers for conversations and entities, allowing Claude to reference historical project details, resolve ambiguous pronouns or references, and maintain a running timeline of related interactions over time. By executing semantic vector search and entity resolution tools behind the scenes, MCP Memory Server allows users to build long-term knowledge graphs without manually re-entering past conversation notes or setup specifications across separate Claude Desktop chat windows.

Category: AI Memory & Context

Tags: cosmosdb, memory, openai, persistence, rag

Visit MCP Memory Server

How to install and configure MCP Memory Server

1. Clone the repository: bash git clone https://github.com/PlumyCat/mcp-memory-server.git cd mcp-memory-server 2. Install dependencies and compile the TypeScript project: bash npm install npm run build 3. Copy the sample environment file and add your credentials: bash cp .env.example .env Provide values for COSMOS_ENDPOINT, COSMOS_KEY, COSMOS_DATABASE_NAME, COSMOS_CONTAINER_CONVERSATIONS, COSMOS_CONTAINER_ENTITIES, and OPENAI_API_KEY. 4. Register the server in Claude Desktop by editing your configuration file: json { "mcpServers": { "memory": { "command": "node", "args": ["/path/to/mcp-memory-server/dist/index.js"], "cwd": "/path/to/mcp-memory-server" } } } 5. Restart Claude Desktop to enable the memory tools.

What you can do with MCP Memory Server

• Recalling project architecture details and previous technical decisions across multiple Claude Desktop sessions without manual prompts. * Resolving conversational ambiguous references such as pronouns or mentions of specific tools using the entity resolution tool. * Conducting semantic searches over past discussions to retrieve relevant solutions and configuration steps from historical logs. * Tracking timeline-based interactions and mentions of specific organizations, libraries, or individuals across stored conversational history. * Automatically extracting and persisting structured entities like software tools, project names, and companies directly into Azure Cosmos DB.

Key facts

• https://github.com/PlumyCat/mcp-memory-server
• AI Memory & Context, Databases & Data Stores
• cosmosdb, memory, openai, persistence, rag

Part of MCP Servers

What is MCP Memory Server?

MCP Memory Server is an open-source Model Context Protocol server that enables persistent memory for Claude Desktop. It extracts entities, generates OpenAI embeddings, and stores conversation data inside Azure Cosmos DB to provide semantic recall across separate user sessions.

How do I install MCP Memory Server?

To install it, clone the GitHub repository, install dependencies with npm install, and run npm run build. Configure your Azure Cosmos DB credentials and OpenAI API key in a .env file, then add the compiled dist/index.js script to your Claude Desktop configuration file under mcpServers.

What prerequisites are required to use MCP Memory Server?

You need Node.js 18 or higher, an active Azure Cosmos DB account configured with dedicated containers, an OpenAI API key for generating vector embeddings, and an MCP-compatible host such as Claude Desktop.

What tools does MCP Memory Server provide?

The server exposes six core tools: memory_store for persisting extracted entity information, memory_search for semantic search, context_inject for retrieving contextual query history, entity_resolve for mapping conversational references, conversation_analyze for pattern metrics, and memory_timeline for entity tracking.

Is MCP Memory Server open source?

Yes, MCP Memory Server is open source and licensed under the MIT License. The code and configuration guides are available in the PlumyCat/mcp-memory-server repository on GitHub.