AWS Emulator: a free local AWS clone written entirely by AI agents

It is built from 438 AWS API models and covers the interfaces of hundreds of services, though how much is actually implemented varies by service and operation. AWS publishes a fidelity score, which measures how closely the emulator behaves like the real thing, with every release on GitHub. The tests behind it live in a separate repository called aws-emulator-evaluation-suite.

What is AWS Emulator for?

The intended uses are local testing, running integration tests in CI/CD pipelines without touching real AWS resources, and prototyping AWS apps with AI coding agents. For those agents there is an "aws-emulator" skill that handles installing, starting and connecting the emulator.

On Linux you install it from an RPM or Debian package, for x86_64 or aarch64. On macOS and Windows you build an image from the project's Dockerfile and run it in Docker. The source code is on GitHub under the CC0 license.

Is AWS Emulator safe to use in production?

No. It is not meant for production. IAM policies and resource-based policies are not evaluated the way AWS evaluates them, so AWS tells people not to use it for security testing. All resources sit in memory and vanish when you stop the emulator.

How the AI agents built it

Senior principal engineer Matthew Goldberg led the work. According to AWS, it began as a weekend experiment, inspired by an Anthropic researcher who had AI agents build a C compiler this year. Goldberg told Kiro, AWS's AI development environment, to write tests based on the AWS CLI, fix them against real AWS until they passed, then build an in-memory implementation that passes them. The first service, Amazon SQS, took about 15 minutes.

Now three kinds of agent teams run in parallel: one writes tests from the API models and checks them on real AWS, one runs those tests against the emulator and reports bugs, and one fixes the bugs. Hundreds of agents work at once and fix thousands of issues a day. The emulator updates several times daily.