Memgraph AI Toolkit

What is Memgraph AI Toolkit about?

The toolkit offers core Python utilities for querying and analyzing Memgraph, a LangChain integration that exposes Memgraph operations as tools, an MCP server implementation for lightweight STDIO communication, and an experimental LLM‑powered agent that automates MySQL‑to‑Memgraph migrations.

How to use Memgraph AI Toolkit?

1. Start a Memgraph instance (e.g., with Docker) enabling schema info.
2. Install any package in editable mode using uv pip install -e <package-path>[test].
3. Run the corresponding test suite with pytest -s to verify the installation.
4. For the migration agent, configure .env (MySQL and Memgraph connection details) and execute uv run main.py inside integrations/agents.
5. Use the LangChain or MCP packages in your Python code to interact with Memgraph through toolkits or STDIO protocols.

Key Features

• Core Toolbox: CLI and Python helpers for direct Memgraph queries and analysis.
• LangChain Integration: Exposes Memgraph operations as LangChain tools, enabling LLM‑driven workflows.
• MCP Server: Lightweight STDIO‑based Model Context Protocol server for tool communication.
• Migration Agent (Experimental):
  • Automatic schema analysis of MySQL databases.
  • LLM‑generated graph modeling with interactive refinement.
  • End‑to‑end data migration, validation, and rollback.
  • Real‑time progress monitoring and detailed logging.

Use Cases

• Building AI‑assisted recommendation or fraud‑detection systems on graph data.
• Integrating LLMs with graph queries via LangChain for natural‑language data exploration.
• Automating legacy relational database migrations to Memgraph for graph‑centric analytics.
• Rapid prototyping of graph‑based AI agents that require fast access to Memgraph via MCP.

FAQ

Q: What Python version is required? A: The packages are compatible with Python 3.9+ and rely on uv for environment management.

Q: Do I need an OpenAI API key? A: Only the LangChain integration tests require OPENAI_API_KEY; runtime usage depends on the LLMs you choose.

Q: Can I run the MCP server on Windows? A: Yes, the Python implementation works on any OS where Memgraph and Python are available.

Q: Is the migration agent production‑ready? A: It is marked experimental; use it for evaluation and provide feedback before production deployment.

Q: How do I contribute? A: Fork the repository, make changes in the relevant subdirectory, and submit a pull request following the contribution guidelines in the repo.

Memgraph AI Toolkit

A unified mono-repo for integrating AI-powered graph tools on top of Memgraph.
This repository contains the following libraries:

1. memgraph-toolbox Core Python utilities and CLI tools for querying and analyzing a Memgraph database. The package is available on the PyPi

2. langchain-memgraph A LangChain integration package, exposing Memgraph operations as LangChain tools and toolkits. The package is available on the PyPi

3. mcp-memgraph An MCP (Model Context Protocol) server implementation, exposing Memgraph tools over a lightweight STDIO protocol. The package is available on the PyPi

4. agents ⚡ Experimental An intelligent database migration agent that automates the process of migrating from MySQL to Memgraph using LLM-powered graph modeling and analysis. Features automated schema analysis, intelligent graph modeling with interactive refinement, and data migration with validation.

Key Features

Migration Agent Capabilities

• 🔍 Intelligent Schema Analysis: Automatically analyzes MySQL database structure and relationships
• 🧠 LLM-Powered Graph Modeling: Uses AI to create optimal graph models from relational schemas
• 🔄 Interactive Refinement: Allows users to refine graph models through natural language feedback
• ⚡ Automated Migration: Handles complete data migration with validation and rollback capabilities
• 📊 Progress Monitoring: Real-time migration progress tracking with detailed logging
• 🛡️ Data Validation: Comprehensive pre and post-migration data integrity checks

Usage examples

For individual examples on how to use the toolbox, LangChain, MCP, or agents, refer to our documentation:

• Langchain examples
• MCP examples
• Agent migration examples - Interactive MySQL to Memgraph migration workflows

Developing locally

You can build and test each package directly from your repo. First, start a Memgraph MAGE instance with schema info enabled:

docker run -p 7687:7687 \
  --name memgraph \
  memgraph/memgraph-mage:latest \
  --schema-info-enabled=true

Once Memgraph is running, install any package in “editable” mode and run its test suite. For example, to test the core toolbox:

uv pip install -e memgraph-toolbox[test]
pytest -s memgraph-toolbox/src/memgraph_toolbox/tests

Core tests

To test the core toolbox, just run:

uv pip install -e memgraph-toolbox[test]
pytest -s memgraph-toolbox/src/memgraph_toolbox/tests

Langchain integration tests

To run the LangChain tests, create a .env file with your OPENAI_API_KEY, as the tests depend on LLM calls:

uv pip install -e integrations/langchain-memgraph[test]
pytest -s integrations/langchain-memgraph/tests

MCP integration tests

uv pip install -e integrations/mcp-memgraph[test]
pytest -s integrations/mcp-memgraph/tests

Agent integration tests

uv pip install -e integrations/agents[test]
pytest -s integrations/agents/tests

To run a complete migration workflow with the agent:

cd integrations/agents
uv run main.py

Note: The agent requires both MySQL and Memgraph connections. Set up your environment variables in .env based on .env.example.

If you are running any test on MacOS in zsh, add "" to the command:

uv pip install -e memgraph-toolbox"[test]"