Memory Service

The Memory Service MCP server provides a persistent, long-term memory to your AI assistant, like Claude Desktop and Cursor.

It goes beyond simple chat history by using semantic search to understand the meaning of the information you store.

The most impressive part is its autonomous, “dream-inspired” consolidation system. This feature automatically organizes and compresses your memories over time to create a self-evolving knowledge base without you having to lift a finger.

Features

• 🧠 Autonomous Memory Consolidation: Inspired by human sleep, it automatically organizes, connects, and compresses memories over time.
• 💾 Dual Storage Backends: Choose between the full-featured ChromaDB or the lightweight and fast SQLite-vec. The installer even recommends one based on your hardware.
• 🗣️ Claude Code Commands: Interact directly with the memory using simple commands like /memory-store and /memory-recall right in your chat.
• 🤝 Multi-Client Coordination: Share a single memory database across multiple applications like Claude Desktop and VS Code without data conflicts.
• 🌐 Zero-Configuration Networking: Uses mDNS for automatic service discovery on your local network, so there’s no need to manually configure endpoints.
• 🔒 Persistent & Private: All your data is stored locally. You have full control over your information.

Use Cases

• Long-Term Project Development: You can store architectural decisions, project requirements, and code snippets. Weeks later, you can ask, “What did we decide about the database authentication?” and get a precise answer. This saves a ton of time trying to find old conversations or notes.
• Personal Knowledge Management: Use it as a personal brain to store ideas, book summaries, and important facts. Because it uses semantic search, you can retrieve information based on concepts, not just keywords. For example, you could ask, “What were those thoughts I had about lean manufacturing?” and it would find the relevant notes.
• Learning & Research: When researching a new topic, you can save key findings, articles, and insights. The memory service can then help you connect disparate pieces of information, revealing relationships you might have missed.

How To Use It

1. Clone the repository and use the installer script. It detects your hardware (Apple Silicon, Intel, Windows, etc.) and optimizes the setup for you.

git clone https://github.com/doobidoo/mcp-memory-service.git 
cd mcp-memory-service

2. Create and activate a Python virtual environment. This keeps its dependencies isolated from your other projects.

# Create the environment 
python -m venv venv 

# Activate it (Mac/Linux) 
source venv/bin/activate 

# On Windows, use: 
venv\Scripts\activate

3. Run the intelligent installer:

bash python install.py

4. After installation, you need to tell Claude Desktop where to find the memory service.

{
  "memory": {
    "command": "python",
    "args": ["-m", "mcp_memory_service.server"],
    "cwd": "/path/to/your/mcp-memory-service",
    "env": {
      "MCP_MEMORY_STORAGE_BACKEND": "sqlite_vec",
      "MCP_MEMORY_SQLITE_PATH": "/path/to/your/mcp-memory/sqlite_vec.db"
    }
  }
}

Make sure to replace the paths with the actual locations on your machine.

5. Once it’s running, you interact with it through natural language in Claude or by using the direct commands.

Store:

"Remember this: the project's main API endpoint is /api/v2/users."

Retrieve:

"What was the main API endpoint I told you about?"

Commands:

claude /memory-store "The staging server is deployed at staging.example.com" --tags "devops,urls"
claude /memory-recall "what did I say about the staging server?"

FAQs

Q: What is this “dream consolidation” feature? Do I need to manage it?
A: It’s an autonomous background process inspired by how the human brain consolidates memories during sleep. It organizes your stored information over days, weeks, and months, finding connections and compressing data to keep it efficient. You don’t need to do anything; it’s enabled by default and manages itself.

Q: Can I use this with other tools besides Claude Desktop?
A: Yes. The server supports multi-client access, allowing you to share the same memory database across different MCP-compatible applications like VS Code (with an MCP extension), Cursor, and others. The installer can help you configure this universal access.

Q: What’s the difference between the ChromaDB and SQLite-vec backends?
A: ChromaDB is more powerful and suited for very large memory collections, offering advanced vector search capabilities. SQLite-vec is a lightweight, faster alternative that’s great for most personal use cases, especially on older hardware, as it uses significantly fewer resources. The installer intelligently picks the best one for your system, but you can override its choice.

Q: I’m on Windows and the installation failed with a PyTorch error. What should I do?
A: This is a common issue on Windows. The project includes a dedicated script to handle this. After activating your virtual environment, run python scripts/install_windows.py. It will detect your system’s capabilities (like CUDA for GPU acceleration) and install the correct PyTorch version.