Memory Vault

Overview

Memory Vault is a Model Context Protocol (MCP) server: Local-first AI memory layer with hybrid search. Postgres + pgvector. Self-hosted, MIT. As an MCP server it plugs straight into AI assistants, exposing its functionality as tools the model can invoke on your behalf.

It runs locally: the Docker image ghcr.io/mihaibuilds/memory-vault-mcp:1.4.0 speaks MCP over stdio on your machine. Configuration happens through environment variables — you'll need DB_HOST, DB_PORT, DB_NAME (among others). The current release is v1.4.0, published under the MIT license, with source at MihaiBuilds/memory-vault on GitHub, where it has earned 60 stars.

You'll find Memory Vault in the Memory category here on mcp.site; it connects to Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Zed, and any other MCP client — see the install snippets below. If you maintain Memory Vault for MihaiBuilds, claim this listing to verify ownership, earn the Verified badge, and keep the details current.

Installation

// claude_desktop_config.json
{
  "mcpServers": {
    "memory_vault": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/mihaibuilds/memory-vault-mcp:1.4.0"
      ],
      "env": {
        "DB_HOST": "<db-host>",
        "DB_PORT": "<db-port>",
        "DB_NAME": "<db-name>",
        "DB_USER": "<db-user>"
      }
    }
  }
}

Requires DB_HOST, DB_PORT, DB_NAME