neo4j-mcp-workspace-template

A template for end-to-end Neo4j graph development using MCP servers. Compatible with Claude Code, Cursor, Gemini CLI, GitHub Copilot (VS Code), OpenCode, OpenAI Codex CLI, and Mistral Vibe.

Quick Start

git clone https://github.com/neo4j-field/neo4j-mcp-workspace-template
cd neo4j-mcp-workspace-template
chmod +x setup.sh && ./setup.sh

Then open your AI coding tool:

setup.sh will:

1. Prompt for your Neo4j credentials and OpenAI API key
2. Write .env with your configuration
3. Install dependencies for the local MCP servers
4. Generate MCP config files for all 7 tools:
   • .mcp.json — Claude Code
   • .cursor/mcp.json — Cursor
   • .gemini/settings.json — Gemini CLI
   • .vscode/mcp.json — GitHub Copilot (VS Code)
   • opencode.json — OpenCode
   • .codex/config.toml — OpenAI Codex CLI
   • .vibe/config.toml — Mistral Vibe

---

Claude Desktop

If you use Claude Desktop (not one of the 7 coding tools above), install the workspace as a Desktop Extension — no git clone needed.

Prerequisites

1. uv — must be installed before the extension can start (see Installing uv below)

2. Two Neo4j databases — the workspace uses a Documents DB (chunks, entities) and an Ontology DB (extraction schema). Two options:

   Option	How
   Two Aura instances	Create two free AuraDB instances — one for documents, one for the ontology. Each has its own URI, username, and password.
   Neo4j Desktop (local)	Install Neo4j Desktop, create one DBMS, then add two databases inside it (e.g. neo4j and ontology). Both share the same URI and credentials.

   Aura free tier note: each Aura instance has a single database named neo4j — you cannot create additional databases. Use two separate instances.

3. An OpenAI API key — for embeddings and entity extraction (or configure another provider via Extra Environment Variables in the install dialog)

Install

1. Download neo4j-mcp-workspace.dxt from the latest release