On September 16, 2026, AWS announced a new getting-started experience for new users that reduces initial setup so they can begin developing sooner. After signing up with an existing identity such as Google or GitHub, users get a project prepared by AWS and can move straight on to building with a coding agent. The experience is being rolled out gradually to new customers, and paid plans also include a per-project spend limit. However, this limit works by stopping the project when it is reached. AWS reduces the effort needed to start building, but users are still left to decide how much cost they can accept and when they can tolerate a stop.
From automatic project creation to building with AI
The "project" that AWS sets up automatically includes an AWS account in which to create resources, along with settings for sharing it within a team. AWS also applies additional security controls, making it easier for users to move from preparing an environment to developing an application. AWS's announcement emphasizes that users can start with appropriate initial settings, in contrast to the traditional procedure of finishing configuration work first.
The background is how AWS has added features over the years to meet the demands of large enterprises and government agencies. Micah Walter, the senior solutions architect who wrote the announcement, explained that global-scale deployment and a broad feature set are necessary for large customers, but every individual setting can be a burden for developers who want to turn a new idea into something real. On September 18, The Register reported that AWS recognized new users find the existing console difficult, and that it had worked on this overhaul for about a year.
The entry point for collaboration changes as well. Team members are invited by email address, and an invited person can access only the designated project. Users do not need to create IAM users individually, so they can add participants without first learning IAM and IAM Identity Center, the mechanisms AWS uses to manage access permissions. In addition, permissions between supported services and resources are configured automatically through console operations or by the coding agent.
What is automated here is the permission setup for supported combinations. AWS is not promising that any configuration on any service can be handled automatically. Even so, when trying out a small application, this reduces the burden of setting up shared use and connections between resources from scratch.
Development with AI begins by pasting the prompt shown after sign-in into your own coding agent. In AWS's demonstration, the agent set up the AWS CLI and the Agent Toolkit for AWS, and created CLAUDE.md, which describes working guidelines for the new environment. When it was then instructed to build an API that returns a unique sequential ID for each request, it reportedly created and deployed a Lambda function, a DynamoDB table, and an API Gateway API.
Walter said the demonstration did not require manually configuring permissions between resources. This is a build example presented by AWS, and it does not prove development time or configuration correctness for every application. The concrete change in the new experience is that preparing the environment to hand to the agent is built into the steps that follow AWS sign-up.
Spend limits include stopping
The spend limit on paid plans controls monthly pre-tax costs on a per-project basis. It is not a fixed fee under a pricing plan; it is a mechanism that holds down a project's pre-tax costs in exchange for stopping it. AWS's announcement says limits start from $20 per month, but the official spend limit guide says the lowest value you can set is the greater of $20 and a conservatively estimated usage forecast.
In other words, a project running many resources cannot always be given a $20 limit. The forecast uses current-month usage and running resources, and heavy usage in the previous month can also affect the minimum. When choosing a limit, the mechanism takes into account not only the amount you want to pay but also the cost of the environment already running.
Comparing the September announcement against the current official guide, the behavior of the spend limit, from setup through to after a stop, breaks down as follows. This applies to paid plans in the new getting-started experience, and is distinct from the optional, earlier cost controls that users can enable.
| Stage | Behavior or condition described in the official guide | Points to consider when using it |
|---|---|---|
| Setting the limit | The minimum is the greater of $20 and a conservatively estimated usage forecast | It is neither a fixed $20 monthly fee nor a limit you can always choose |
| Reaching the limit | The project and all its resources are stopped; data is retained | Staying under the cost cap takes priority over continued operation |
| Resuming use | Raise the limit in AWS Settings; some resources require a manual restart | Raising the limit does not necessarily restore all functions automatically |
| Leaving it stopped | If no action is taken within 90 days of the stop, data is permanently deleted | It is not a way to store data indefinitely while stopped |
This breakdown shows that a feature meant to control costs also directly affects whether the application runs. AWS itself positions the spend limit for experimentation, learning, and sandboxes. Using it in production requires being able to tolerate a short stop if unexpected costs occur. For a developer who wants to keep prototyping with peace of mind and an operator running a service that cannot go down, the same limit amount means different things.
Notifications arrive before a stop occurs. They are sent when actual costs reach 50%, 75%, and 90% of the limit, and when costs are projected to reach or exceed the limit within the next 10 days. Users can review the limit at that point, but receiving a notification and being able to accept a stop are separate judgments.
Free tier conditions and the path to advanced features
Under the new experience, most new customers can sign up without a credit card and receive $100 in AWS Free Tier credits. However, as the official sign-up documentation shows, payment information may be requested in some cases, and some users may fall outside the free tier. In the latter case, the user registers payment information and starts on a paid plan. The no-card explanation cannot be read as a condition common to all users.
Availability is also limited for now, as a phased rollout to a limited set of new customers. The new experience may not yet appear on the sign-up screen, and this is not an announcement that all existing accounts will switch at once. Whether you can use it and which plan you start on need to be checked separately.
For the path after growth, AWS has provided the option of enabling advanced features. Users who need multiple Regions, or who want to set their own management policies in AWS Organizations, can enable them at no additional charge, AWS says. AWS explains that users can move to a configured Organizations environment without migration work or downtime, and that their earlier settings are kept. What is free here is enabling the advanced features, not the charges for the resources you use.
When deciding whether to use the new entry point, the key considerations are whether the services you need are supported, and whether you can accept the stop that occurs when the cost limit is reached. If those conditions fit a prototype or learning project, you can start learning AWS from a working application without first spending time on permissions and environment preparation.