Neon MCP Server

Neon MCP Server is an MCP server that connects large language models to Neon serverless Postgres databases via the Neon API. Developers, data engineers, and non-technical team members use it to inspect, query, and manage relational database environments through plain conversational prompts. By establishing a bridge between MCP clients and Neon management endpoints, the server enables users to perform complex administrative tasks without writing manual API calls or switching between separate database dashboards. Supported capabilities include creating projects, provisioning compute endpoints, generating and altering database branches, querying table schemas, and executing database migrations. Users can also enforce security parameters by scoping operations to specific projects, filtering available tool categories, or locking execution into a read-only mode that restricts destructive operations. Designed primarily for local development workflows and IDE integrations, the server facilitates conversational schema inspection, data browsing, and multi-branch testing workflows inside environments like Cursor, VS Code, and Claude Code.

Category: Cloud & Infrastructure

Tags: database-management, neon, postgres, SQL

Visit Neon MCP Server

How to install and configure Neon MCP Server

1. Run automated initialization for Cursor, VS Code, or Claude Code using the command npx neon@latest init. 2. Alternatively, register the remote hosted server globally across supported agents using npx add-mcp "https://mcp.neon.tech/mcp?category=projects&category=branches&category=endpoints&category=querying&category=schema". 3. To configure manually in your MCP client config (such as mcp.json), add a server entry named Neon with type set to http and url set to https://mcp.neon.tech/mcp?category=projects&category=branches&category=endpoints&category=querying&category=schema. 4. If using API key authentication instead of browser OAuth, include the headers field with "Authorization": "Bearer <$NEON_API_KEY>". 5. Restart your MCP client and complete the browser OAuth prompt if prompted.

What you can do with Neon MCP Server

• Provisioning new Neon Postgres projects and configuring compute endpoints directly through conversational prompts in an IDE. * Branching a database to test migrations and schema changes in isolation before deploying them to parent branches. * Running read-only SQL queries against Neon Postgres tables to inspect rows and debug application data. * Inspecting database schemas, table structures, and relationships without writing manual information schema SQL queries. * Summarizing active projects and branch hierarchies across personal or organizational accounts from Claude Code or Cursor.

Key facts

• https://github.com/neondatabase-labs/mcp-server-neon
• Cloud & Infrastructure, Databases & Data Stores
• database-management, neon, postgres, SQL

Part of MCP Servers

How do I install Neon MCP Server?

You can set up the server by executing npx neon@latest init in your terminal to automatically configure your editor, or by running npx add-mcp with the hosted Neon endpoint. You can also manually add the remote endpoint URL https://mcp.neon.tech/mcp directly to your client mcp.json configuration file.

What can Neon MCP Server do?

The server translates natural language requests into Neon API operations and SQL executions. It allows you to create projects, provision branches, inspect table schemas, run database migrations, view endpoint metrics, and execute queries directly from conversational AI interfaces.

Which MCP clients work with Neon MCP Server?

Neon MCP Server works with any client supporting the Model Context Protocol over remote HTTP connections, including Cursor, Claude Code, GitHub Copilot in VS Code, and Kiro.

Is Neon MCP Server open source?

Yes, Neon MCP Server is open source and licensed under the MIT License on GitHub.

Can I restrict Neon MCP Server to read-only mode?

Yes. You can enable read-only mode by deselecting full access during OAuth authorization, or by appending the query parameter ?readonly=true to the hosted server URL. In read-only mode, write actions like migrations, branch creations, and destructive SQL queries are blocked.