We've shipped a hosted version of our MCP server with OAuth

Changelog

The latest product updates from Neon

RSS feed

Neon MCP Server in the cloud

We’ve brought the Neon MCP Server to the cloud. Our hosted MCP server makes it easier to integrate AI workflows into clients like Cursor, Windsurf, and Claude Desktop—no API keys or local setup required.

You can start using it today by pointing your client to:

https://mcp.neon.tech

How to try it with Cursor:

  1. Open Cursor Settings

  2. Under MCP Servers, add:

    "Neon": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.neon.tech/sse"]
    }

That’s it—you’re connected to Neon’s remote MCP Server.

We’re releasing this in preview while the MCP OAuth spec continues to evolve. Things might change, and we’d love your feedback as we improve.

📖 Read the full announcement for more info and a demo video.

New safeguards for protected branches

We added a warning and confirmation modal to the SQL Editor when running queries on protected branches. This helps prevent accidental changes to production data. You'll see a clear notice and must confirm before proceeding.

SQL Editor warnings for protected branches

Create Neon projects directly from the Azure Portal

For users of Neon on Azure — you can now create Neon projects directly from the Azure Portal. Creating a project is part of Neon Serverless Postgres resource creation. You can also add Neon projects to an existing Neon resource from a new Projects page. All Neon plans, including the Free plan, support creating multiple Neon projects.

Azure project form

Fixes & improvements
  • Neon Console

    The Computes tab on individual branch pages in the Neon Console now shows Started and Suspended labels for the primary compute, indicating when the compute was last started or suspended.

    compute started label

  • Neon API

    We added a started_at attribute to the Retrieve compute endpoint details response. This timestamp shows when your Neon compute was last started.

  • Vercel Native Integration

    New Neon projects (referred to as Databases in Vercel) now use Postgres 17 by default. Previously, projects created through the Vercel Native Integration used Postgres 15.

  • Neon App for Slack

    We've added a new /neon disconnect command to the Neon App for Slack. This command lets you remove your Neon account connection and unsubscribe from all channels while keeping the app installed for future use. You can use it when you need to switch accounts or temporarily pause notifications.

    As a reminder, you can use /neon subscribe in any channel to start receiving notifications again. The bot will guide you through any necessary setup steps.

    To install the app or learn more about all available commands, see Neon App for Slack.

  • Fixes

    Fixed an issue that caused the Tables page in the Neon Console to reload when the browser page regained focus.

Neon serverless driver is now generally available (GA)

The Neon serverless driver for JavaScript/TypeScript has reached version 1.0.0 and is now generally available! Built for environments like Vercel Functions, Cloudflare Workers, and even browsers, the driver carries SQL over HTTP or WebSockets — no raw TCP required.

This GA release brings a cleaner, more maintainable codebase, stronger SQL injection safeguards, support for composable tagged-template queries, and better performance when inserting binary data over HTTP.

What you need to know about the 1.0.0 GA release:

  • It requires Node.js v19 or later.

  • It includes a breaking change but only if you're calling the HTTP query template function as a conventional function. The first usage shown below remains safe and supported. However, the second usage is an SQL injection risk (notice the parentheses) and is no longer permitted and now throws an error. You'll need to update your app if you use it.

    // this usage remains safe and supported
    const resultA = await sql`SELECT * FROM table WHERE id = ${id}`;
    // this usage is not safe and now throws an error
    const resultB = await sql(`SELECT * FROM table WHERE id = ${id}`);

For more about these changes and others in the 1.0.0 GA release, please see the 1.0.0 release notes or read the blog post.

Streamlined Neon Auth setup

With Neon Auth, your user data is available right in your database.

We've simplified Neon Auth onboarding so you can add authentication to your project faster. With a cleaner UI and clearer steps, it's now easier to get started adding your first users with Neon Auth.

Streamlined Neon Auth setup

Learn more in our docs.

MCP Server updates

We're continuing to improve our MCP Server. This week, we added built-in support for the AI rules we announced two weeks ago. In case you missed it, we provide AI rules for setting up Neon Auth, querying your database with the Neon serverless driver, and integrating Neon with Drizzle ORM. Keep an eye out for more AI rules as we expand this resource.

Windows developers will also find improved platform compatibility in this release. See the MCP Server changelog to find the latest updates.

Fixes & improvements
  • Getting started panel

    Added a new Integrate Neon with your AI tools option to the Project Dashboard, making it easier to connect Neon with AI tools like Cursor, Windsurf, Zep, Qdrant, and Weaviate.

    new ai card in get started panel

  • Neon API

    • Added started_at timestamp field to the Endpoint response object. This field indicates when a compute endpoint was last started, providing better visibility into compute lifecycle events.
    • Updated the Delete VPC endpoint API to clarify that deleted VPC endpoints cannot be re-added to the same Neon organization.
  • Drizzle Studio update

    We updated the Drizzle Studio integration that powers the Tables page in the Neon Console to version 1.0.19. For the latest improvements and fixes, see the Neon Drizzle Studio Integration Changelog.

Search faster with pg_search on Neon

pg_search extension

We’ve partnered with ParadeDB to bring pg_search to Neon, delivering up to 1,000x faster full-text search inside Postgres. The pg_search extension introduces Elasticsearch-grade ranking, typo tolerance, faceted search, and JSON-aware filtering—all within Postgres. Powered by a BM25-based inverted index, pg_search outperforms native Postgres FTS (tsvector/tsquery), making it ideal for search-heavy applications. It’s available now for Neon databases running Postgres 17 on AWS. To learn more, read the announcement and check out our pg_search docs.

Neon spend is now MACC-eligible on Azure

Neon Postgres purchases made through the Azure Marketplace are now counted toward your Microsoft Azure Consumption Commitment (MACC). As an Azure Benefit Eligible partner, any spend on Neon Postgres via the Azure Marketplace helps fulfill your committed Azure spend with no extra steps required. Learn more.

Get usage notifications in Slack

You can now receive Neon usage notifications directly in your Slack channels! Get updates on your resource usage, find your projects, and invite team members to your organization - without leaving your workspace.

Neon Slack commands including new subscribe feature

Use /neon subscribe in any public channel to start receiving notifications, and /neon unsubscribe to turn them off. The bot will guide you through any necessary setup steps.

To learn more, see the Neon App for Slack docs.

Fixes & improvements
  • Neon Console

    • Expanded the database drop-down menu width in the Neon SQL Editor to accommodate longer database names. Previously, longer names were not fully visible due to the narrow menu width.
    • Added an Unable to fetch projects message to the Projects page in the Neon Console. Previously, an error page was displayed when the project list couldn't be retrieved.
  • Autoscaling default settings

    We've updated the default autoscaling settings for newly created projects on paid Neon plans to provide a better balance of performance and efficiency:

    Neon planMinimum compute sizeMaximum compute size
    Launch14
    Scale18
    Business18

    These optimized defaults help ensure projects scale smoothly to meet workload demands while maintaining cost efficiency. This change applies only to newly created projects—existing projects and computes remain unaffected. You can review and adjust your autoscaling settings anytime in your project settings. From your Project Dashboard, go to Settings > Compute.

  • Postgres effective_cache_size setting is now optimized for better query plans

    Previously, Neon didn’t explicitly set the Postgres effective_cache_size Postgres parameter, so it defaulted to 4 GiB—often too low for larger compute sizes and autoscaling configurations. We now set this value based on the maximum size of Neon's Local File Cache (LFC) for a compute's maximum compute size, which helps the Postgres query planner make better decisions and improves query performance. For information about maximum LFC size per compute size, see the table in How to size your compute.

  • Neon API

  • Drizzle Studio update

    We updated the Drizzle Studio integration that powers the Tables page in the Neon Console to version 1.0.18. For the latest improvements and fixes, see the Neon Drizzle Studio Integration Changelog.

  • Fixes

    • Resolved an issue where resetting a role password in the Neon Console would result in an "invalid password" error in the SQL Editor and on the Tables page.
    • Revised the copy at the bottom of the Connect to your database modal for older Neon projects. The copy previously mentioned that passwords are stored, which is only true for Neon projects created after password storage was introduced.
Was this page helpful?

Subscribe to our changelog.
No spam, guaranteed.