Changelog

Now you can better understand how your projects are consuming credits when using Netlify’s AI Gateway and/or Agent Runners

For all credit-based plans, you can now find an AI inference usage graph in your Usage and Billing dashboard.

AI inference is a usage meter that measures the costs of using AI models and agents on Netlify. This usage meter calculates costs for using Agent Runners and AI Gateway.

The new AI inference usage graph shows AI provider requests and tokens used across your Netlify team’s projects over time.

To check it out, go to Usage and Billing > Account usage insights > AI inference.

You can also review the docs to learn more about monitoring usage and costs for AI features.

Note that Netlify’s AI Gateway and Agent Runners are only available for credit-based plans.

Starting today, you can purchase credit packs directly from your Netlify Dashboard.

Only Team Owners with a credit-based Personal or Pro plan can purchase credit packs.

Plan	Credits per pack	Price
Personal	500	$5 USD
Credit-based Pro	1,500	$10 USD

This gives you more control over your spending and allows you to add credits on demand without setting up auto-recharge.

Previously, you had to configure auto-recharge to purchase additional credits. Auto-recharge is still available, but you now have the flexibility to buy credit packs at any time.

For example, if you’re building new functionality using Agent Runners, you can purchase credit packs to cover the additional usage without enabling auto-recharge. You may still want to use auto-recharge, but now you have the flexibility to manage spending and credit usage in more ways.

Learn more about buying credit packs in our documentation.

React Router 7 recently launched its long-awaited middleware feature, stable since v7.9.0.

Middleware is now fully supported on Netlify.

How to use it

Ensure you’re on version 7.9.0 or later of React Router:

npm install react-router@latest @react-router/node@latest @react-router/dev@latest

Upgrade to version 2 of the Netlify React Router plugin:

npm install @netlify/vite-plugin-react-router@latest

After that, middleware works out of the box.

Next steps

• Check out the full React Router middleware docs
• Create a new React Router 7 project with our updated template
• Read the Netlify React Router plugin docs for more on loader/action context and to access type-safe Netlify context in middleware

We’ve released skew protection for all plans starting with these frameworks:

• Astro 5.15.0+
• Next.js

This feature prevents version mismatch errors when users interact with your site during a deployment.

What it fixes

When you deploy a new version of your site, users with an open browser tab may still be running the old version. If they navigate to a new page or trigger a request, the browser expects a previous version of your application, but the server returns the new version.

This version skew causes:

• Broken assets and 404 errors
• JavaScript errors
• Degraded user experience

How it works

Skew protection pins each user session to a specific Netlify deploy, ensuring the browser and server stay synchronized throughout that session.

To learn how it works more in-depth, check out our skew protection blog post.

Next steps

• Set up skew protection for your framework
• Ensure your framework can support skew protection with Netlify using our Frameworks API

The Next.js 16 release is here and Netlify is ready.

Deploy the latest version, including Turbopack, React Compiler, and enhanced caching APIs, with zero configuration.

• Turbopack (stable): Up to 10× faster Fast Refresh and 2-5× faster builds
• React Compiler (stable): Automatic memoization for smoother re-renders
• Improved caching APIs: revalidateTag with cacheLife profile argument for Stale-While-Revalidate behavior and Server Actions-only updateTag to immediately refresh cached data

Template update

We’ve updated the Next Platform Starter to v16, showcasing Turbopack and the new caching APIs in action. In our tests, the project saw a 2.8× faster next build with Turbopack.

Upgrading to v16 on Netlify

If you’re upgrading from v15, no Netlify configuration changes are required. Teams that have opted out of automatic updates for Netlify’s OpenNext adapter will need to manually upgrade, or simply opt back in to receive automatic updates.

Resources

• Deploy the demo
• Read the Next.js 16 release
• View Next.js on Netlify docs

Now you can require all production deploys to go through a Git-based workflow.

When enabled in your project settings, this setting prevents:

• Accidental production publishes from the Netlify CLI, MCP server, or API
• Promoting Deploy Previews or branch deploys to production through the Netlify UI

This ensures that both team members and AI agents follow your preferred Git-based workflow before your project’s code reaches production.

Learn more in our Git workflow docs.

Claude Haiku 4.5 is now available through Netlify’s AI Gateway — no extra setup required.

You can use the native Anthropic SDK without managing API keys or external accounts.

Here’s how to use it today in a Netlify Function with the Anthropic SDK:

import Anthropic from "@anthropic-ai/sdk";

export default async () => {
	const anthropic = new Anthropic();

	const response = await anthropic.messages.create({
		model: "claude-haiku-4-5-20251001",
		messages: [
			{
				role: "user",
				content: "Give me pros and cons of using claude-haiku-4-5-20251001 over other models."
			},
		],
	});
}

This is supported across Edge, Background, and Scheduled Functions and includes features like Rate-Limiting, access to Netlify’s advanced caching primitives, and many others.

Learn more in the AI Gateway docs.

GPT-5 Pro is now available through Netlify’s AI Gateway — no extra setup required. You can use the OpenAI SDK directly, without needing to manage API keys or external accounts.

Here’s how you can use it today in Netlify Functions:

import { OpenAI } from "openai";

export default async () => {
  const openai = new OpenAI();
  const response = await openai.responses.create({
    model: "gpt-5-pro",
    input: "Give pros and cons of using GPT-5 Pro over open models.",
  });
};

This is supported across Edge, Background, and Scheduled Functions and includes features like Rate-Limiting, access to Netlify’s advanced caching primitives, and many others.

See more in the AI Gateway docs.

To support new AI workflows and our AI inference usage meter we’re rebalancing our credit-based pricing plans:

Add-on credit rates (effective immediately)

• Personal: 200 credits for $5 → 500 credits for $5
• Pro: 1,000 credits for $20 → 1,500 credits for $10

These new rates apply to all credit-based plans when you auto-recharge credits.

Pro plan monthly credits (effective next billing cycle)

• Before: 5,000 credits/month
• After: 3,000 credits/month

To smooth the transition, anyone who signed up for a Pro credit-based plan prior to October 1st will receive a 2,000 credit bonus in their October billing cycle.

View plans or estimate costs with our pricing calculator.