Getting started with the GraphAcademy MCP server

Ask your coding assistant to teach you Neo4j and it will do its best from training data - explanations that may be out of date, exercises against a database it cannot see, and no record of what you have already covered. It has no idea which courses exist, what you have completed, or whether the Cypher you just wrote actually produced the right graph.

The GraphAcademy MCP server fixes that. It is a remote Model Context Protocol server that gives any MCP-compatible client access to GraphAcademy courses, your enrolments and progress, a managed Neo4j database, and a set of Neo4j code-generation tools.

This guide assumes you already use one of these clients and have a GraphAcademy account. If you don't have an account, sign up first.

Add the server to your client

The server runs remotely at https://mcp.graphacademy.neo4j.com/mcp, so there is nothing to install or run locally - you only add a config entry.

claude mcp add --transport http \
  graphacademy \
  https://mcp.graphacademy.neo4j.com/mcp \
  --scope user

Snippets for Codex CLI, Gemini CLI, and Windsurf are available in your GraphAcademy account page.

The first time your client calls a tool, it opens a browser window and asks you to sign in with your GraphAcademy account. The server uses standard OAuth, so the sign-in is handled by the client - you never paste an API key or token. Once authorised, everything you do through the server is tied to your account: a course you complete in your editor shows as complete on the website, and vice versa.

Take a course from your editor

The server registers slash commands in clients that support MCP prompts. Enrol in a course by slug:

/enrol course_slug: neo4j-fundamentals

Enrolling does more than flip a flag. If the course needs a database, the server provisions a personal Neo4j instance for you and can hand your agent the credentials - ask for them and the get_database_credentials tool returns a ready-to-paste .env block.

From there, /next-lesson fetches the next incomplete lesson with its content and exercises, and your agent teaches it in the conversation. When a lesson ends with a hands-on challenge, complete it and run /verify - the server executes the lesson's verification Cypher against your own database and only marks the lesson complete when every check passes. Your agent cannot wave you through; the graph has to be right.

Between lessons, the run_cypher tool lets your agent (or you, through it) query the course database directly, so you can explore the data, test a hypothesis, or debug a failed verification without opening another tool.

In clients like ChatGPT that support embedded MCP apps, open_cypher_workspace goes one further and renders an interactive query editor inline in the conversation.

Generate Neo4j code for your own project

The second half of the toolset is not about courses at all. Five generation tools cover the lifecycle of a Neo4j application:

• neo4j_graph_modeler designs a graph data model - labels, relationships, properties, indexes, and constraints - from a plain-language description of your domain.
• neo4j_import_advisor recommends an import strategy and generates the scripts, whether that is LOAD CSV or neo4j-admin.
• neo4j_mock_data_generator produces realistic seed data generation code in JavaScript, TypeScript, Python, Java, or Go.
• neo4j_query_builder writes Cypher for you, and can introspect a live schema to get it right.
• neo4j_project_builder scaffolds application code: driver setup, session management, and parameterised queries.

The /build-neo4j-app command chains them into a guided workflow - describe what you want to build and your agent works through model, mock data, queries, and application code in order.

Pair it with the Neo4j MCP server

The GraphAcademy server teaches and generates; the official Neo4j MCP server connects your agent to a live database of your own for schema inspection and query execution. They are complementary: design the model here, run it there.

The /neo4j-stack command explains the recommended setup, and the Developing with Neo4j MCP Tools course covers it in depth.

Where to go next

Add the server, run /enrol with a course that interests you, and see how far you can get without leaving your editor. And if using MCP has you curious about how servers like this are built, there is a course for that too.

Building GraphRAG Python MCP tools

Build a custom MCP server in Python with FastMCP. Create Neo4j graph-backed tools, resources, and prompts that ground AI agents with GraphRAG.

View course