AWS Emulator: local AWS APIs, written entirely by AI agents

What it is

AWS Emulator is a local program that serves AWS APIs over HTTPS at 127.0.0.1:9297. The AWS CLI, AWS SDKs and coding agents talk to it with the same calls they use against the cloud, redirected with environment variables (AWS News Blog).

• No additional service fee and no AWS account needed. You pay only for AWS resources when you deploy.
• Installs as an RPM or Debian package on Linux and runs in Docker on macOS and Windows. An embedded mode exists for Rust tests.
• State lives in memory. Each instance gets its own port, keys and state, so parallel agents and CI jobs never see each other's resources.
• Unsupported operations return 501 Not Implemented.

Built by agents, unreviewed by humans

AWS Senior Principal Engineer Matthew Goldberg says his team has not written or reviewed any of the code (About Amazon). Three teams of agents do the work: one writes fidelity tests from the AWS API models and verifies them against real AWS, one replays the tests to find divergence, and one fixes each issue. A merge gate checks that changes compile and pass agent reviews for coding standards, security and fidelity regressions.

The project started from an empty server stub for 438 AWS API models. It now has tens of thousands of fidelity tests, and thousands of issues get fixed every day by hundreds of concurrent agents. It began as a weekend experiment with Kiro.

Caveats

Don't use it for security testing: it does not evaluate IAM and resource-based policies the way AWS does, so a request AWS would deny may succeed. It sends anonymous usage data by default, which you can turn off. Keep real credentials out of your test environment.

Why it matters

Two reasons. Coding agents get feedback in milliseconds without a deployment, which makes iteration faster and cheaper. And it is a working pattern for building a large system with agents from a spec and a reference implementation: the fidelity tests and the merge gate do the heavy lifting, not the model.