Neo4j MCP Server
A Model Context Protocol server that enables running Cypher queries and retrieving schema from Neo4j databases, supporting AI-driven graph exploration and text-to-Cypher workflows.
README
🔍⁉️ Neo4j MCP Server
🌟 Overview
A Model Context Protocol (MCP) server implementation that provides database interaction and allows graph exploration capabilities through Neo4j. This server enables running Cypher graph queries, analyzing complex domain data, and automatically generating business insights that can be enhanced further with an application's analysis tools.
This MCP server facilitates Text2Cypher workflows like the one detailed below.
- Blue steps are handled by the agent
- Purple by the Cypher or another MCP server
- Green by the user
A user question is input to the process and the output is an answer generated by the agent.

🧩 Components
🛠️ Tools
The server offers these core tools:
📊 Query Tools
-
read_neo4j_cypher- Execute Cypher read queries to read data from the database
- Input:
query(string): The Cypher query to executeparams(dictionary, optional): Parameters to pass to the Cypher query
- Returns: Query results as JSON serialized array of objects
- Timeout: Read queries are subject to a configurable timeout (default: 30 seconds) to prevent long-running queries from disrupting conversational flow
-
write_neo4j_cypher- Execute updating Cypher queries
- Input:
query(string): The Cypher update queryparams(dictionary, optional): Parameters to pass to the Cypher query
- Returns: A JSON serialized result summary counter with
{ nodes_updated: number, relationships_created: number, ... }
🕸️ Schema Tools
get_neo4j_schema- Get a list of all nodes types in the graph database, their attributes with name, type and relationships to other node types
- No input required
- Returns: JSON serialized list of node labels with two dictionaries: one for attributes and one for relationships
🏷️ Namespacing
The server supports namespacing to allow multiple Neo4j MCP servers to be used simultaneously. When a namespace is provided, all tool names are prefixed with the namespace followed by a hyphen (e.g., mydb-read_neo4j_cypher).
This is useful when you need to connect to multiple Neo4j databases or instances from the same session.
⚙️ Query Configuration
The server provides configuration options to optimize query performance and manage response sizes:
⏱️ Query Timeouts
Configure timeouts for read queries to prevent long-running queries from disrupting conversational flow:
Command Line:
mcp-neo4j-cypher --read-timeout 60 # 60 seconds
Environment Variable:
export NEO4J_READ_TIMEOUT=60
Docker:
docker run -e NEO4J_READ_TIMEOUT=60 mcp-neo4j-cypher:latest
Default: 30 seconds. Read queries that exceed this timeout will be automatically cancelled to maintain responsive interactions with AI models.
📏 Token Limits
Control the maximum size of query responses to prevent overwhelming the AI model:
Command Line:
mcp-neo4j-cypher --token-limit 4000
Environment Variable:
export NEO4J_RESPONSE_TOKEN_LIMIT=4000
Docker:
docker run -e NEO4J_RESPONSE_TOKEN_LIMIT=4000 mcp-neo4j-cypher:latest
When a response exceeds the token limit, it will be automatically truncated to fit within the specified limit using tiktoken. This ensures:
- Consistent Performance: Responses stay within model context limits
- Cost Control: Prevents excessive token usage in AI interactions
- Reliability: Large datasets don't break the conversation flow
Note: Token limits only apply to read_neo4j_cypher responses. Schema queries and write operations return summary information and are not affected.
🏗️ Local Development & Deployment
🐳 Local Docker Development
Build and run locally for testing or remote deployment:
# Build the Docker image with a custom name from your local version of the server
docker build -t mcp-neo4j-cypher:latest .
# Run locally (uses http transport by default for Docker)
docker run -p 8000:8000 \
-e NEO4J_URI="bolt://host.docker.internal:7687" \
-e NEO4J_USERNAME="neo4j" \
-e NEO4J_PASSWORD="your-password" \
mcp-neo4j-cypher:latest
# Access the server at http://localhost:8000/api/mcp/
🚀 Transport Modes
The server supports different transport protocols depending on your deployment:
- STDIO (for local development): Standard input/output for Claude Desktop and local tools
- HTTP (for remote deployments): RESTful HTTP for web deployments and microservices
- SSE: Server-Sent Events for legacy web-based deployments
Choose your transport based on use case:
- Local development/Claude Desktop: Use
stdio - Remote deployment: Use
http - Legacy web clients: Use
sse
🔒 Security Protection
The server includes comprehensive security protection with secure defaults that protect against common web-based attacks while preserving full MCP functionality when using HTTP transport.
🛡️ DNS Rebinding Protection
TrustedHost Middleware validates Host headers to prevent DNS rebinding attacks:
Secure by Default:
- Only
localhostand127.0.0.1hosts are allowed by default - Malicious websites cannot trick browsers into accessing your local server
Environment Variable:
export NEO4J_MCP_SERVER_ALLOWED_HOSTS="example.com,www.example.com"
🌐 CORS Protection
Cross-Origin Resource Sharing (CORS) protection blocks browser-based requests by default:
Environment Variable:
export NEO4J_MCP_SERVER_ALLOW_ORIGINS="https://example.com,https://example.com"
🔧 Complete Security Configuration
Development Setup:
mcp-neo4j-cypher --transport http \
--allowed-hosts "localhost,127.0.0.1" \
--allow-origins "http://localhost:3000"
Production Setup:
mcp-neo4j-cypher --transport http \
--allowed-hosts "example.com,www.example.com" \
--allow-origins "https://example.com,https://example.com"
🚨 Security Best Practices
For allow_origins:
- Be specific:
["https://example.com", "https://example.com"] - Never use
"*"in production with credentials - Use HTTPS origins in production
For allowed_hosts:
- Include your actual domain:
["example.com", "www.example.com"] - Include localhost only for development
- Never use
"*"unless you understand the risks
🔧 Usage with Claude Desktop
Using DXT
Download the latest .dxt file from the releases page and install it with your MCP client.
💾 Released Package
Can be found on PyPi https://pypi.org/project/mcp-neo4j-cypher/
Add the server to your claude_desktop_config.json with the database connection configuration through environment variables. You may also specify the transport method, namespace and other config variables with cli arguments or environment variables.
{
"mcpServers": {
"neo4j-database": {
"command": "uvx",
"args": [ "mcp-neo4j-cypher@0.4.0", "--transport", "stdio" ],
"env": {
"NEO4J_URI": "bolt://localhost:7687",
"NEO4J_USERNAME": "neo4j",
"NEO4J_PASSWORD": "<your-password>",
"NEO4J_DATABASE": "neo4j"
}
}
}
}
🌐 HTTP Transport Configuration
For custom HTTP configurations with security middleware:
# Complete HTTP configuration with security
mcp-neo4j-cypher --transport http \
--server-host 127.0.0.1 \
--server-port 8080 \
--server-path /api/mcp/ \
--allowed-hosts "localhost,127.0.0.1,example.com" \
--allow-origins "https://yourapp.com"
# Using environment variables
export NEO4J_TRANSPORT=http
export NEO4J_MCP_SERVER_HOST=127.0.0.1
export NEO4J_MCP_SERVER_PORT=8080
export NEO4J_MCP_SERVER_PATH=/api/mcp/
export NEO4J_MCP_SERVER_ALLOWED_HOSTS="localhost,127.0.0.1,example.com"
export NEO4J_MCP_SERVER_ALLOW_ORIGINS="https://yourapp.com"
mcp-neo4j-cypher
Multiple Database Example
Here's an example of connecting to multiple Neo4j databases using namespaces:
{
"mcpServers": {
"movies-neo4j": {
"command": "uvx",
"args": ["mcp-neo4j-cypher@0.4.0", "--namespace", "movies"],
"env": {
"NEO4J_URI": "neo4j+s://demo.neo4jlabs.com",
"NEO4J_USERNAME": "recommendations",
"NEO4J_PASSWORD": "recommendations",
"NEO4J_DATABASE": "recommendations"
}
},
"local-neo4j": {
"command": "uvx",
"args": ["mcp-neo4j-cypher@0.4.0"],
"env": {
"NEO4J_URI": "bolt://localhost:7687",
"NEO4J_USERNAME": "neo4j",
"NEO4J_PASSWORD": "password",
"NEO4J_DATABASE": "neo4j",
"NEO4J_NAMESPACE": "local"
}
}
}
}
In this setup:
- The movies database tools will be prefixed with
movies-(e.g.,movies-read_neo4j_cypher) - The local database tools will be prefixed with
local-(e.g.,local-get_neo4j_schema)
Syntax with --db-url, --username, --password, --read-timeout and other command line arguments is still supported but environment variables are preferred:
<details> <summary>Legacy Syntax</summary>
"mcpServers": {
"neo4j": {
"command": "uvx",
"args": [
"mcp-neo4j-cypher@0.4.0",
"--db-url",
"bolt://localhost",
"--username",
"neo4j",
"--password",
"<your-password>",
"--namespace",
"mydb",
"--transport",
"sse",
"--server-host",
"127.0.0.1",
"--server-port",
"8000"
"--server-path",
"/api/mcp/"
]
}
}
</details>
🐳 Using with Docker
Here we use the Docker Hub hosted Cypher MCP server image with stdio transport for use with Claude Desktop.
Config details:
-i: Interactive mode - keeps STDIN open for stdio transport communication--rm: Automatically remove container when it exits (cleanup)-p 8000:8000: Port mapping - maps host port 8000 to container port 8000NEO4J_TRANSPORT=stdio: Uses stdio transport for Claude Desktop compatibilityNEO4J_NAMESPACE=neo4j: Prefixes tools with "neo4j-" (e.g.,neo4j-read_neo4j_cypher)NEO4J_URI=bolt://host.docker.internal:7687: Allows Docker container to connect to Neo4j running on host machine
{
"mcpServers": {
"neo4j": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-p",
"8000:8000",
"-e", "NEO4J_URI=bolt://host.docker.internal:7687",
"-e", "NEO4J_USERNAME=neo4j",
"-e", "NEO4J_PASSWORD=password",
"-e", "NEO4J_NAMESPACE=neo4j",
"-e", "NEO4J_TRANSPORT=stdio",
"mcp/neo4j-cypher:latest"
]
}
}
}
🐳 Docker Deployment
The Neo4j MCP server can be deployed using Docker for remote deployments. Docker deployment should use HTTP transport for web accessibility. In order to integrate this deployment with applications like Claude Desktop, you will have to use a proxy in your MCP configuration such as mcp-remote.
📦 Using Your Built Image
After building locally with docker build -t mcp-neo4j-cypher:latest .:
# Run with http transport (default for Docker)
docker run --rm -p 8000:8000 \
-e NEO4J_URI="bolt://host.docker.internal:7687" \
-e NEO4J_USERNAME="neo4j" \
-e NEO4J_PASSWORD="password" \
-e NEO4J_DATABASE="neo4j" \
-e NEO4J_TRANSPORT="http" \
-e NEO4J_MCP_SERVER_HOST="0.0.0.0" \
-e NEO4J_MCP_SERVER_PORT="8000" \
-e NEO4J_MCP_SERVER_PATH="/mcp/" \
mcp/neo4j-cypher:latest
# Run with security middleware for production
docker run --rm -p 8000:8000 \
-e NEO4J_URI="bolt://host.docker.internal:7687" \
-e NEO4J_USERNAME="neo4j" \
-e NEO4J_PASSWORD="password" \
-e NEO4J_DATABASE="neo4j" \
-e NEO4J_TRANSPORT="http" \
-e NEO4J_MCP_SERVER_HOST="0.0.0.0" \
-e NEO4J_MCP_SERVER_PORT="8000" \
-e NEO4J_MCP_SERVER_PATH="/mcp/" \
-e NEO4J_MCP_SERVER_ALLOWED_HOSTS="example.com,www.example.com" \
-e NEO4J_MCP_SERVER_ALLOW_ORIGINS="https://example.com" \
mcp/neo4j-cypher:latest
🔧 Environment Variables
| Variable | Default | Description |
|---|---|---|
NEO4J_URI |
bolt://localhost:7687 |
Neo4j connection URI |
NEO4J_USERNAME |
neo4j |
Neo4j username |
NEO4J_PASSWORD |
password |
Neo4j password |
NEO4J_DATABASE |
neo4j |
Neo4j database name |
NEO4J_TRANSPORT |
stdio (local), http (remote) |
Transport protocol (stdio, http, or sse) |
NEO4J_NAMESPACE |
(empty) | Tool namespace prefix |
NEO4J_MCP_SERVER_HOST |
127.0.0.1 (local) |
Host to bind to |
NEO4J_MCP_SERVER_PORT |
8000 |
Port for HTTP/SSE transport |
NEO4J_MCP_SERVER_PATH |
/api/mcp/ |
Path for accessing MCP server |
NEO4J_MCP_SERVER_ALLOW_ORIGINS |
(empty - secure by default) | Comma-separated list of allowed CORS origins |
NEO4J_MCP_SERVER_ALLOWED_HOSTS |
localhost,127.0.0.1 |
Comma-separated list of allowed hosts (DNS rebinding protection) |
NEO4J_RESPONSE_TOKEN_LIMIT |
(none) | Maximum tokens for read query responses |
NEO4J_READ_TIMEOUT |
30 |
Timeout in seconds for read queries |
🌐 SSE Transport for Legacy Web Access
When using SSE transport (for legacy web clients), the server exposes an HTTP endpoint:
# Start the server with SSE transport
docker run -d -p 8000:8000 \
-e NEO4J_URI="neo4j+s://demo.neo4jlabs.com" \
-e NEO4J_USERNAME="recommendations" \
-e NEO4J_PASSWORD="recommendations" \
-e NEO4J_DATABASE="recommendations" \
-e NEO4J_TRANSPORT="sse" \
-e NEO4J_MCP_SERVER_HOST="0.0.0.0" \
-e NEO4J_MCP_SERVER_PORT="8000" \
--name neo4j-mcp-server \
mcp-neo4j-cypher:latest
# Test the SSE endpoint
curl http://localhost:8000/sse
# Use with MCP Inspector
npx @modelcontextprotocol/inspector http://localhost:8000/sse
🐳 Docker Compose
For more complex deployments, you may use Docker Compose:
version: '3.8'
services:
# Deploy Neo4j Database (optional)
neo4j:
image: neo4j:5.26.1 # or another version
environment:
- NEO4J_AUTH=neo4j/password
- NEO4J_PLUGINS=["apoc"]
ports:
- '7474:7474' # HTTP
- '7687:7687' # Bolt
volumes:
- neo4j_data:/data
# Deploy Cypher MCP Server
mcp-neo4j-cypher-server:
image: mcp/neo4j-cypher:latest
ports:
- '8000:8000'
environment:
- NEO4J_URI=bolt://host.docker.internal:7687
- NEO4J_USERNAME=neo4j
- NEO4J_PASSWORD=password
- NEO4J_DATABASE=neo4j
- NEO4J_TRANSPORT=http
- NEO4J_MCP_SERVER_HOST=0.0.0.0 # must be 0.0.0.0 for sse or http transport in Docker
- NEO4J_MCP_SERVER_PORT=8000
- NEO4J_MCP_SERVER_PATH=/api/mcp/
- NEO4J_NAMESPACE=local
depends_on:
- neo4j
volumes:
neo4j_data:
Run with: docker-compose up -d
🔗 Claude Desktop Integration with Docker
For Claude Desktop integration with a Dockerized server using http transport:
{
"mcpServers": {
"neo4j-docker": {
"command": "npx",
"args": ["-y", "mcp-remote@latest", "http://localhost:8000/api/mcp/"]
}
}
}
Note: First start your Docker container with HTTP transport, then Claude Desktop can connect to it via the HTTP endpoint and proxy server like mcp-remote.
🚀 Development
📦 Prerequisites
- Install
uv:
# Using pip
pip install uv
# Using Homebrew on macOS
brew install uv
# Using cargo (Rust package manager)
cargo install uv
- Clone the repository and set up development environment:
# Clone the repository
git clone https://github.com/neo4j-contrib/mcp-neo4j.git
cd mcp-neo4j-cypher
# Create and activate virtual environment using uv
uv venv
source .venv/bin/activate # On Unix/macOS
.venv\Scripts\activate # On Windows
# Install dependencies including dev dependencies
uv pip install -e ".[dev]"
- Run Integration Tests
./tests.sh
🔧 Development Configuration
For development with Claude Desktop using the local source:
{
"mcpServers": {
"neo4j-dev": {
"command": "uv",
"args": ["--directory", "/path/to/mcp-neo4j-cypher", "run", "mcp-neo4j-cypher", "--transport", "stdio", "--namespace", "dev"],
"env": {
"NEO4J_URI": "bolt://localhost:7687",
"NEO4J_USERNAME": "neo4j",
"NEO4J_PASSWORD": "<your-password>",
"NEO4J_DATABASE": "neo4j"
}
}
}
}
Replace /path/to/mcp-neo4j-cypher with your actual project directory path.
📄 License
This MCP server is licensed under the MIT License. This means you are free to use, modify, and distribute the software, subject to the terms and conditions of the MIT License. For more details, please see the LICENSE file in the project repository.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。