Vertex AI Memory Bank MCP Server
Provides tools for generating, retrieving, and managing long-term memories using Vertex AI Memory Bank, enabling LLMs to store and recall user context and preferences.
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., "@Vertex AI Memory Bank MCP Serverremember that I prefer dark mode"
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.
This is a personal project by Ivan Nardini to explore how to build a Model Context Protocol (MCP) server for Vertex AI Memory Bank.
Vertex AI Memory Bank MCP server is not a Google product. And it is not officially support.
Vertex AI Memory Bank MCP Server
A simple MCP (Model Context Protocol) server that enables LLMs to generate and retrieve long-term memories using Vertex AI's Memory Bank.
Why This Project?
This server demonstrates how to build an MCP server with Vertex AI Memory Bank. It has been inspired by a developer request and released for developers.
Related MCP server: MCP-Mem0
Prerequisites
Python 3.11 or higher
Google Cloud account with Vertex AI API enabled
Basic understanding of async Python (helpful but not required)
Quick Start
Setup Google Cloud
# Install gcloud CLI (if not already installed)
# https://cloud.google.com/sdk/docs/install
# Authenticate
gcloud auth application-default login
# Set your project
gcloud config set project YOUR_PROJECT_ID
# Enable Vertex AI API
gcloud services enable aiplatform.googleapis.comInstall
# Clone the repository
git clone https://github.com/yourusername/vertex-ai-memory-bank-mcp.git
cd vertex-ai-memory-bank-mcp
# Install with pip
pip install -r requirements.txt
# OR install with uv (faster, recommended)
uv sync
# For running examples (optional)
pip install -e ".[examples]"
# OR with uv
uv sync --extra examplesConfigure
# Copy the example environment file
cp .env.example .env
# Edit .env with your project details
GOOGLE_CLOUD_PROJECT=your-project-id
GOOGLE_CLOUD_LOCATION=us-central1Run Your First Example
Interactive Tutorial (Recommended): Open get_started_with_memory_bank_mcp.ipynb in Jupyter
Or try the command-line examples:
# Basic MCP Client Usage
python examples/basic_usage.py
# Gemini Agent with Memory
python examples/gemini_memory_agent.py
# Automatic Tool Calling with Gemini
python examples/automatic_tool_calling.pyUse with Claude Desktop
Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"memory-bank": {
"command": "python",
"args": ["/path/to/memory_bank_server.py"],
"env": {
"GOOGLE_CLOUD_PROJECT": "your-project-id",
"GOOGLE_CLOUD_LOCATION": "us-central1"
}
}
}
}Key Concepts
Memory Scope
Memories are scoped to users or contexts:
scope = {"user_id": "alice123"}Memory Topics
Categorize what to remember:
topics = ["USER_PREFERENCES", "USER_PERSONAL_INFO"]Semantic Search
Find relevant memories with similarity search:
search_query = "programming preferences"
top_k = 5Available Tools
Tool | Purpose | Example Use Case |
| Set up connection to Vertex AI | First-time setup |
| Extract memories from conversations | After chat sessions |
| Fetch relevant memories | Personalize responses |
| Manually add a memory | Store user preferences |
| Remove specific memory | User requests deletion |
| View all stored memories | Debugging/inspection |
Common Patterns
Pattern 1: Conversation Memory
# After each conversation turn
await session.call_tool(
"generate_memories",
{
"conversation": conversation_history,
"scope": {"user_id": user_id},
"wait_for_completion": True
}
)Pattern 2: Explicit Memory
# Store specific facts
await session.call_tool(
"create_memory",
{
"fact": "User prefers dark mode",
"scope": {"user_id": user_id}
}
)Pattern 3: Context Retrieval
# Get relevant context before responding
memories = await session.call_tool(
"retrieve_memories",
{
"scope": {"user_id": user_id},
"search_query": user_message,
"top_k": 5
}
)Project Structure
vertex-ai-memory-bank-mcp/
├── memory_bank_server.py # Main entry point
├── src/ # Modular source code
│ ├── __init__.py
│ ├── server.py # Server orchestration
│ ├── tools.py # MCP tool implementations
│ ├── config.py # Configuration management
│ ├── app_state.py # Application state
│ ├── validators.py # Input validation
│ └── formatters.py # Data formatting
├── examples/ # Usage examples
│ ├── basic_usage.py # Basic MCP client usage
│ ├── automatic_tool_calling.py # Automatic function calling
│ └── claude_config.json # Claude Desktop config
├── get_started_with_memory_bank_mcp.ipynb # Getting started tutorial
├── pyproject.toml # Project config (pip & uv)
├── requirements.txt # Dependencies (pip)
├── uv.lock # Lock file (uv)
├── .env.example # Environment template
├── .gitignore # Git ignore rules
├── .python-version # Python version
├── README.md # This file
└── LICENSE # Apache 2.0 LicenseTroubleshooting
"Connection closed" error
Solution: Check that your MCP server is using stderr for logging, not stdout.
"Not authenticated"
Solution: Run gcloud auth application-default login
Contributing
This project is meant to inspire. Feel free to fork and create your own version as well as share your production implementations.
Resources
Interactive Tutorial - Start here!
License
This project is licensed under the Apache 2.0 License.
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Latest Blog Posts
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/inardini/vertex-memory-bank-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server