AWS Shipped a Free Cloud Emulator Without Writing Its Code
AI ProductsNovember 12, 20265 min readBy Digitalaikart Editorial

AWS Shipped a Free Cloud Emulator Without Writing Its Code

#AIProducts#AWS#AIAgents#IndiaAI

TL;DR

01.What AWS actually shipped

On 11 November 2026, AWS announced AWS Emulator, a local emulator of AWS services that runs on your own machine. The AWS CLI, the SDKs and coding agents talk to it over a local HTTPS endpoint using the same APIs as the cloud, so application code does not change. You install it from an RPM or Debian package on Linux, or run it in Docker on macOS and Windows.

It is written in Rust so that it stays light on a laptop, and its services are generated from the same AWS API models that back the SDKs, so requests and error responses have the same shape as on real AWS. It covers more than 200 AWS services at launch, GeekWire reported. Unsupported operations return 501 Not Implemented, and each copy gets its own port, keys and in-memory state, so parallel test runs do not collide.

The terms are the striking part. AWS says there is no additional service fee, that you pay only for resources you actually deploy, and that the source sits on GitHub under the CC0 1.0 public domain dedication.

02.No human wrote or reviewed the code

AWS Emulator was written entirely by autonomous AI agents. In a post on Amazon's newsroom, Senior Principal Engineer Matt Goldberg wrote that his team had not written or reviewed any of the code, and that the guardrails around the agents were what made that possible. Simon Willison noted the project logged 29,684 commits in its first three weeks.

The build runs as three teams of agents. One writes fidelity tests, short programs that run a workflow against real AWS and assert exactly how it behaves, down to status codes and error messages. A second replays those tests, finds where the emulator diverges and files an issue. A third fixes one issue at a time and must clear a merge gate checking compilation, coding standards, security and fidelity regressions. AWS says hundreds of agents work concurrently and thousands of issues are fixed daily.

It began as a weekend experiment. Goldberg asked Kiro, AWS's coding agent, to write tests from the AWS CLI, iterate until they passed against real AWS, then build an in-memory implementation those same tests would pass. Kiro returned a workable Amazon SQS in about 15 minutes; the rest followed over a weekend. The trigger was Anthropic's earlier demonstration that agents could build a C compiler, and AWS had the two ingredients that made it repeatable: a published spec in its API models, and a reference implementation.

03.Why this lands well in India

For a small Indian team, test environments are a line item paid in dollars. A DynamoDB table, an SQS queue and a Lambda function spun up to check one code path costs money every time a developer or a CI job needs it, and a careless CI configuration can quietly run all night. A local emulator that answers in milliseconds and bills nothing removes most of that, which matters more when the bill arrives in dollars and the budget is in rupees.

There is a second benefit in how Indian developers now work. An agent that must deploy to a region to check whether its change works is slow and expensive; one that can create a queue, send a message and read it back on the laptop iterates in seconds.

Running locally also keeps test data on the machine rather than in a cloud account, which is convenient for teams working with customer records, payment flows or KYC documents. That is a convenience, not a compliance guarantee.

04.The limits you must respect

AWS is direct about what the emulator does not do. It does not evaluate IAM policies the way AWS does, so a request real AWS would deny may succeed locally. Each emulator acts as a single account with no cross-account or cross-Region behaviour. Accuracy varies by service, and AWS publishes a fidelity score per service, with the full test suite public so you can run it yourself.

The instruction that follows is blunt: do not use the emulator to test whether your application is secure. Keep real credentials out of test job environments, and test security settings against real AWS before production. For a fintech or health startup that is not a formality.

There is also the honesty question that hangs over any agent-built codebase. AWS's claim that no human reviewed the code rests on its own account of the process. What a reader can check is the output, which is why the published fidelity tests matter more than the headline.

05.What to do with it this week

If you maintain a service that touches SQS, DynamoDB or S3, the cheapest experiment is to point your existing tests at the emulator. Install it, start it, and set the environment variables so the SDK talks to the local endpoint; code and tests stay as they are. Whatever breaks tells you either that your tests depend on real-cloud behaviour you had not noticed, or that coverage is thin. Both are useful.

Check the fidelity report for the services you use before trusting a green run, and keep a real-AWS integration stage for anything touching permissions, encryption or billing. For students and early-career developers it is a free way to practise cloud patterns on a laptop, without an AWS account and without watching a credit balance.

The larger signal is worth carrying into planning meetings. A hyperscaler has now shipped a public tool whose code no human at the company wrote, with its test suite in the open so the claim can be checked. That is a different proposition from a demo.

Frequently asked questions

Is AWS Emulator free to use?
Yes. AWS says the emulator carries no additional service fee and that you pay only for AWS resources when you deploy on AWS. The source code is published on GitHub under the CC0 1.0 public domain dedication. You install it from an RPM or Debian package on Linux, or run it in Docker on macOS and Windows.

Can I trust tests that pass on AWS Emulator?
Partly. AWS measures fidelity with tests verified against real AWS and publishes a score for each service as a number of passing tests out of the total, so you can check the services you rely on and run the public suite yourself. But the emulator does not evaluate IAM policies the way AWS does, acts as a single account with no cross-account or cross-Region behaviour, and its accuracy varies by service and feature. Test security and access control on real AWS before production.

Who wrote the AWS Emulator code?
Autonomous AI agents did. AWS Senior Principal Engineer Matt Goldberg wrote that his team had not written or reviewed any of the code. The work is split across three agent teams, one maintaining fidelity tests, one finding divergences and filing issues, and one fixing them through a merge gate that checks compilation, coding standards, security and fidelity. AWS says hundreds of agents work concurrently, and Simon Willison noted 29,684 commits in the project's first three weeks.

Found this useful? Share it

Source