What Is AWS Emulator
AWS has released AWS Emulator, a local emulator of its APIs. The program runs on a laptop, a development server, or in a continuous integration pipeline, and the platform's official tools, such as the CLI and SDKs, communicate with the emulator instead of the real cloud services: it is enough to set a few environment variables, without modifying the code. Neither an AWS account nor payment for the service is required. According to the documentation, resources remain in memory and are lost when the program is shut down; the DynamoDB example even reproduces the error message that AWS returns for a table name that is too short.
Code Written by Autonomous Agents
The unusual element is who produced the code: according to the source, autonomous agents wrote and reviewed it in its entirety. The project started from a prompt to the Kiro system, which in about fifteen minutes returned a working Amazon SQS emulator. AWS then divided the work among three teams: those who write the tests derive scenarios from the public Smithy models and verify them against the real service; those who look for defects rerun the tests on the emulator and record failures; those who fix take a single issue and propose a change.
Every change passes a check covering compilation, code review, security, and fidelity, and several security agents conduct adversarial reviews. Matthew Goldberg, cited by the source, speaks of hundreds of agents in parallel resolving thousands of issues per day. The release pipeline uses AWS developer tools, with CodeCommit as the agents' repository; the code reaches GitHub only at the end.
Coverage and Limits, to Be Verified
The emulator is built from 438 API models and covers more than 200 services, with one Rust crate per service. The numbers, however, are less reassuring than the label suggests: in version v1.0.322, evaluated on November 11, 2026, it passes 24,243 of the 42,857 classified fidelity tests, or 56.6%. The best service is AWS KMS at 93.9%, followed by Amazon SQS at 89.8%; the worst is CloudFormation at 15.8%.
According to the fidelity documentation, control APIs are the best reproduced, while support for data-plane, CloudFormation, Terraform, and CDK varies from service to service; unimplemented operations return an HTTP 501 error. Professional use therefore requires some caution: the overall percentage says little about how reliable an individual service is, and the tests also include failed ones.
There remains the matter of building from source: it requires a Rust toolchain, a C compiler, and at least 100 GB of free space, and HTTPS on 127.0.0.1 with a self-signed certificate is required. The release pace is fast: versions from v1.0.265 to v1.0.322 were released between November 8 and 11. The source links to the repository, the user guide, the test suite, and an engineering article. It does not, however, document which services teams actually use in test flows, nor how the agents handle edge cases.
