Neon for AI Agents
Use Neon as the Postgres backend for your agents
AI agents can now provision infrastructure, including databases. With AI agents already creating databases every few seconds, they are poised to manage a significant portion of the web's infrastructure in the future — and, like developers, AI agents love working with Neon: Replit partnered with Neon to back Replit Agents, which are already creating thousands of Postgres databases per day. Learn more.
What makes Neon a good database for AI Agents
-
One-second provision times. AI Agents generate code in seconds, so it's a bad user experience to wait minutes for a new Postgres instance to be deployed. Neon provisions databases nearly instantaneously, eliminating this friction.
-
Scale to zero makes empty databases economically feasible. Some databases created by agents might only be used for a few minutes; if you’re the company behind the agent, you’ll quickly have a large database fleet full of inactive databases. With Neon, that’s not a problem—you can still maintain this fleet within a reasonable budget.
-
Straightforward API that even an AI Agent can use. The same API endpoints that are useful for developers managing large database fleets on Neon are also perfect for AI Agents. With the Neon API, you can not only create and delete databases but also track usage, limit resources, and handle configuration.
-
Neon is 100% Postgres. The most-loved database by developers worldwide is also the best choice for AI agents, thanks to its versatility (it works for almost any app) and the vast amount of resources, examples, and training datasets available.
Tools for AI Agents
We recently published a package on NPM called @neondatabase/toolkit, merging the already existing packages into a single SDK that is easier for AI agents to consume. Read more.
With a few lines of code, AI agents can use the Neon toolkit to create a Postgres database on Neon, run SQL queries, and tear down the database. Here's a quick look:
import { NeonToolkit } from "@neondatabase/toolkit";
const toolkit = new NeonToolkit(process.env.NEON_API_KEY!);
const project = await toolkit.createProject();
await toolkit.sql(
project,
`
CREATE TABLE IF NOT EXISTS
users (
id UUID PRIMARY KEY,
name VARCHAR(255) NOT NULL
);
`,
);
await toolkit.sql(
project,
`INSERT INTO users (id, name) VALUES (gen_random_uuid(), 'Sam Smith')`,
);
console.log(await toolkit.sql(project, `SELECT name FROM users`));
await toolkit.deleteProject(project);Neon also supports a Model Context Protocol (MCP) server that lets you use any MCP Client, such as Claude Desktop, to manage Postgres databases with Neon using natural language; for example:
Create a new Postgres database, and call it "my-database". Let's then create a table called users with the following columns: id, name, email, and password.I want to run a migration on my project called "my-project" that alters the users table to add a new column called "created_at".Can you give me a summary of all of my Neon projects and what data is in each one?
Both tools are open source. You can find them on GitHub.