AWS has released AWS Emulator, a local emulator of the AWS APIs. You run it on a laptop, a dev server or in CI. The AWS CLI and SDKs talk to it instead of AWS. No AWS account is needed, and there is no service fee. The unusual part: autonomous agents wrote and reviewed all of its code. For developers and coding agents, AWS Emulator gives a fast, free test loop against AWS-shaped APIs.
TL;DR
- Size: Built from 438 AWS API models, covering 200+ services. One Rust crate per service.
- Runs on: Linux x86_64/aarch64 (RPM or Debian), or Docker on macOS/Windows with a 2 GiB memory limit. Source builds need 100 GB of free disk.
- Performance: Passes 24,243 of 42,857 scored fidelity tests (56.6%) in release v1.0.322.
- Best: AWS KMS at 93.9% (291 of 310 tests); Amazon SQS at 89.8%.
- Worst: AWS CloudFormation at 15.8% (414 of 2,612 tests).
What is AWS Emulator?
AWS Emulator is a headless local program from AWS that serves an AWS API endpoint. You point the CLI or an SDK at it with environment variables. Your code does not change. Resources live in memory and vanish when it stops.
The docs show a DynamoDB example. A CreateTable call with a 2-character name fails with the same ValidationException text AWS returns.
How did AI agents build AWS Emulator?
The project began with a prompt to Kiro. It returned a passable Amazon SQS emulator in about 15 minutes. AWS then split the work across 3 teams of agents:
- Test writers derive scenarios from the public Smithy API models and verify each against real AWS.
- Issue finders replay those tests against the emulator and file failures.
- Fixers take 1 issue each and submit a patch.
Every patch passes a merge gate covering build, code review, security and fidelity. Several security agents run adversarial reviews. Matthew Goldberg reports hundreds of concurrent agents fixing thousands of issues a day. The release pipeline runs on AWS Developer Tools, with CodeCommit as the agents’ repository. Code reaches GitHub last.