Enables natural language question-answering over stored memory data and semantic embeddings using OpenAI's API for enhanced search capabilities.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Memvid MCP Serversearch my project memory for API authentication patterns"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Memvid MCP Server
An Model Context Protocol (MCP) server for Memvid, enabling AI agents to use persistent, file-based memory.
Features
Project-based Memory: Create and manage isolated memory files (
.mv2) for different projects.Add Content: Store text documents, code snippets, and metadata.
Semantic Search: Find relevant content using Memvid's hybrid search (lexical + semantic).
Ask Memory (Optional): Ask natural language questions about your stored data (requires OpenAI API Key).
Installation
npm install -g memvid-mcp-server
# or
bun add -g memvid-mcp-serverConfiguration
This server can be used with any MCP-compliant client (e.g., Claude Desktop, Cursor).
Environment Variables
Variable | Description | Required |
| Your OpenAI API Key. Required to enable the | No (Optional) |
| Set to | No (Optional) |
Claude Desktop Configuration
Add the following to your claude_desktop_config.json:
{
"mcpServers": {
"memvid": {
"command": "npx",
"args": ["-y", "memvid-mcp-server@latest"],
"env": {
"OPENAI_API_KEY": "sk-...",
"MEMVID_LOCAL_STORAGE": "0"
}
}
}
}Note: If you do not provide
OPENAI_API_KEY, theask_memorytool will not be available, but you can still usecreate_or_open_memory,add_content, andsearch_memory(lexical mode).
Tools
create_or_open_memory: Initialize a new project memory or open an existing one.project_name(string): Unique identifier for the project.storage_path(string, optional): Absolute path to the PARENT directory (e.g., project root). The server creates amemvid_mcpfolder inside this path.
add_content: Add text and metadata to the memory.project_name(string): Unique identifier.content(string): Text content to store.storage_path(string, optional): Absolute path to storage directory.
search_memory: Search your memory.project_name(string): Unique identifier.query(string): Search query (use*for all recent items).storage_path(string, optional): Absolute path to storage directory.
ask_memory: (Optional) Ask questions about your memory using an LLM.project_name(string): Unique identifier.question(string): Question to ask.storage_path(string, optional): Absolute path to storage directory.
memvid_delete_project: Delete a project's memory.project_name(string): Unique identifier.storage_path(string, optional): Absolute path to storage directory.
Tip: All tools accept an optional
storage_pathargument. This allows the client to explicitly define where the memory project is stored, overriding the default behavior. Useful for containerized environments.
Development
# Install dependencies
bun install
# Run locally
bun run src/index.ts
# Build
bun build ./src/index.ts --compile --outfile serverCredits
This MCP Server is a wrapper around the powerful Memvid SDK. Full credit goes to the Memvid team for their excellent technology.
NPM Package: @memvid/sdk
Documentation: docs.memvid.com
This server cannot be installed
Resources
Looking for Admin?
Admins can modify the Dockerfile, update the server description, and track usage metrics. If you are the server author, to access the admin panel.