MemoVault

MemoVault exposes an MCP server that lets you manage and query a personal memory store from an external client, such as Claude Code. It provides flexible backends, semantic memory search, and a simple JSON storage option, enabling memory-enabled interactions with your LLM workflows.

How to use

You interact with MemoVault through an MCP client by configuring an MCP server entry and using memory-related commands. The MCP server enables you to store memories, search for relevant memories, chat with memory context, and manage memory records, all while leveraging your chosen LLM backend.

Typical usage workflows include adding memories as you go, querying for context when you need it, and prompting the memory-enabled agent to incorporate past preferences into its responses. You can also persist memories to disk for long-term retention.

How to install

Prerequisites: you need Python and a working environment to install Python packages.

pip install memovault

# For local embeddings (optional)
pip install memovault[local]

After installation, you can start using MemoVault in your Python code or connect via the MCP server configuration described later.

Configuration and MCP server setup

Configure MemoVault to expose an MCP endpoint that Claude Code can connect to. The following example shows how you define the MCP server in your client configuration. This config uses a local process that MemoVault runs as an MCP server.

{
  "mcpServers": {
    "memovault": {
      "command": "memovault-mcp",
      "env": {
        "MEMOVAULT_LLM_BACKEND": "openai",
        "MEMOVAULT_OPENAI_API_KEY": "sk-..."
      }
    }
  }
}

In this setup, MemoVault runs as a local process that accepts MCP commands from the client. The environment variables configure which LLM backend to use and provide authentication for the API if needed.

Usage with Claude Code and memory commands

Once MemoVault is configured as an MCP server, you can issue memory-related commands from the Claude Code interface. Example commands include remembering preferences and asking what MemoVault recalls about you.

Remember that I prefer dark mode
What do you know about my preferences?

Notes on installation defaults

By default MemoVault reads configuration from environment variables or a .env file if you are using Python. You can customize LLM backends, embedder backends, and storage options via environment variables as shown in the configuration section.

Tooling and capabilities overview

MemoVault exposes the following MCP tools to manage and interact with memories.

What you can do with MemoVault

• Store new information in your memory store

• Retrieve memories matching a search term or concept

• Chat with the agent using memory context to inform responses

• Fetch a memory by its unique identifier

• Delete a memory by ID or criteria

• List recently stored memories

• Erase all memories from the store

Security and data handling

Treat stored memories with care. Use proper access controls for your MCP client and protect API keys or credentials used by MemoVault. When using cloud-backed LLM backends, be mindful of data that may be sent to the provider.

Troubleshooting tips

If MemoVault does not appear as an MCP server in the client, verify that the command is accessible in your environment and that the required environment variables are set. Check that the MCP configuration is loaded by your client and that the memory store has the proper read/write permissions.

Notes

This MCP server configuration uses a local process approach. No external HTTP URL is required for the included setup.

Available tools

add_memory

Store new information into MemoVault for future retrieval and context use.

search_memories

Query the memory store to retrieve memories relevant to a given topic or prompt.

chat_with_memory

Interact with the memory-enabled agent using past memories to inform responses.

get_memory

Retrieve a specific memory by its unique identifier.

delete_memory

Remove a memory from the store by ID or criteria.

list_memories

Show recently stored memories in chronological or configured order.

clear_memories

Erase all memories from the store for a fresh start.