AWS Emulator: Amazon Releases a Free Local API Emulator Written and Maintained by AI Agents
On November 11, AWS made AWS Emulator available, a local emulator of its APIs designed to test code and coding agents without an AWS account or deployments. Free, written in Rust, and released into the public domain, it is continuously built and patched by AI agents: over 70 releases between November 8 and 13. It does not evaluate IAM policies.
Testing code that calls AWS has always come with a cost: an account, resources to provision and tear down, and deployments that take minutes. For a coding agent seeking to verify its work in tight feedback loops, this is a major bottleneck. Amazon is addressing this with AWS Emulator, available since November 11: a free local emulator of its APIs that responds to the same commands as the cloud.
How It Works
The emulator is a command-line tool, written in Rust to stay lightweight, that serves an HTTPS endpoint on 127.0.0.1:9297. It generates its own credentials (prefixed with EMULATOR to avoid any confusion with real keys) and rejects requests signed with anything else. Pointing the AWS CLI or an SDK to it takes just a few environment variables, leaving application code untouched. On Linux, it installs via RPM or Debian packages; on macOS and Windows, users must build a Docker image using the Dockerfile provided in the documentation, as AWS does not publish a pre-built one. For tests written in Rust, an embedded mode runs an isolated instance per test without network overhead.
In terms of coverage, AWS does not provide an exhaustive list: the emulator is generated from public AWS API models—the same ones used to produce the SDKs—ensuring requests and errors match the format used by AWS. An operation that is not yet implemented returns a 501 error, causing tests to fail immediately rather than letting them pass mistakenly.
Software Written by Agents—and It Shows
The most unusual aspect is how it is built. According to AWS, the emulator is built and maintained by AI agents: one group compares emulator responses to those from AWS and opens a ticket for each discrepancy, another writes tests and patches, and an independent security agent reviews the changes before they ship to production. Fittingly, the repository does not accept external contributions, only issue reports.
The development pace is unlike any traditional project. The releases page jumps from v1.0.269 on November 8 to v1.0.341 on November 13, representing over 70 releases in five days, each listing dozens of automatically generated fixes. Many are tagged as “stubs,” signaling that an operation was previously just an empty shell. For users, this is a double-edged sword: gaps are filled quickly, but behavior changes multiple times a day. With every release, AWS publishes a fidelity score per service (passing tests out of total existing tests) alongside the test suites themselves in a separate repository, allowing anyone to re-run them.
Limitations to Know Before Relying on It
AWS is explicit about what the tool cannot do. Most importantly, it does not evaluate IAM policies the way AWS does: a request denied in the cloud might succeed in the emulator. As such, it should never be used to validate security configurations. It simulates a single account with no Region isolation; its behavior is only verified in the Oregon Region; it is not designed for prolonged runs or load testing; and it sends anonymous usage telemetry by default, which can be turned off. Finally, AWS acknowledges that fidelity is higher for resource provisioning and management than for data operations, and that it varies when used alongside infrastructure-as-code tools like CloudFormation, Terraform, or the CDK.
Our take
AWS Emulator primarily addresses a pressing need for coding agents: offering them a sandbox where they can spin up a DynamoDB table, test it, and patch their code in seconds without risking a live account or running up a bill. The fact that the tool itself is built by agents also makes it a showcase: over 70 releases in five days demonstrates what automated maintenance looks like at scale, complete with genuine safeguards (public fidelity test suites) as well as a moving target. In practice, it cleanly replaces handcrafted mocks for unit and integration testing—provided teams pin a specific version in CI/CD and retain a testing stage on a real AWS account before production, especially for anything involving permissions.
Sources
- AWS News Blog — Test your applications against AWS locally with AWS Emulator (Daniel Abib, November 11, 2026) Primary source
- AWS What's New — AWS Emulator is now available (November 11, 2026) Primary source
- AWS Emulator — documentation (What is AWS Emulator? and Fidelity) Primary source
- AWS Emulator — documentation, Fidelity in AWS Emulator Primary source
- GitHub — aws/aws-emulator, releases page Primary source
- GitHub — aws/aws-emulator (repository and releases page) Primary source
A primary source comes directly from the company, lab or organization concerned. A secondary source reports the information without being its author: we flag it as such. Spotted an error? See our corrections policy.
