AWS has released an emulator that runs on a developer machine and answers the same API calls as the cloud. It carries no additional service fee, and you pay only for AWS resources when you deploy on AWS. Coding agents may gain the most, because they can loop through create, check and fix cycles without waiting on a deployment in a Region. The limits matter as much as the savings, especially for security testing.

What happened

On 11 November, AWS announced AWS Emulator on its News Blog. It serves HTTPS from the developer’s own machine. The AWS CLI, the AWS SDKs and coding agents reach it through environment variables, so application code stays unchanged. Its source is on GitHub, and AWS describes it as generally available.

Linux users install it from RPM or Debian packages. On macOS and Windows, you build a Docker image from the short Dockerfile in the documentation and run the emulator in a container, because AWS publishes no ready-made image. The emulator is written in Rust and built from the API models the SDKs also use, so AWS says requests and errors have the same shape as on AWS.

For Rust tests there is also an embedded mode. You add the aws-emulator-embedded crate from GitHub to your dev dependencies, and each test then runs its own emulator instance inside the test process with separate state.

SiliconANGLE reported that autonomous AI agents built the tool and that it can replicate hundreds of AWS services. The AWS post is more specific: support covers services that have an API model. AWS says agents hunt for gaps against real AWS, write fixes and review them, and a separate security agent checks changes before release. If the emulator lacks an operation, it returns HTTP 501, so tests fail right away.

Why it matters

Local AWS emulation is not new as a category, but this version comes from AWS itself. Because it is generated from AWS’s own API models and checked against real AWS, it may prove more faithful than a hand-built mock. That is our reading, and only the published scores can confirm it.

The fidelity scoring is the notable design choice. A fidelity test is a short SDK program, verified against real AWS, that passes only if the emulator matches AWS on every response and on the resources created. AWS publishes pass counts per service with each release, as k of n, and keeps the full test set public in a GitHub repository, including tests that fail. Buyers can judge coverage for their own stack instead of trusting a claim of broad support. AWS also warns that a score can fall when new tests are added, so read trends with care.

The agent angle looks like a major driver. Coding agents iterate quickly, while cloud deployments are slow. A local target shortens that loop. This suggests AWS wants to reduce friction for agent-driven development. AWS says only that the emulator has no additional service fee, so we cannot say how it fits AWS’s revenue.

What it means for you

Start with the fidelity scores in the AWS Emulator documentation. List the AWS services your code calls, then check each one before moving any test over. A service with a low score can produce false failures or, worse, false passes.

Keep security testing in the cloud. AWS states the emulator does not evaluate IAM or resource-based policies as AWS does, so a request AWS denies might succeed locally. Permission bugs will slip through unless a run on real AWS still gates release.

Check your other constraints. Each emulator acts as a single account, with no cross-account or cross-Region behavior. That likely makes it a poor fit for testing multi-account designs. Some SDKs, such as the Java one, need a trust store setting for the emulator certificate. Mac and Windows teams need to build the Docker image themselves. By default the emulator also sends anonymous usage and error data, and you can turn that off.

If you run agents against AWS today, consider pointing them at the emulator first and keeping real credentials out of their environment. Our cloud coverage tracks related cost-control moves, such as Google Cloud spend caps, if sandbox bills worry you.

What to watch

  • Fidelity scores for the services you rely on, and whether they rise from release to release.
  • Embedded mode beyond Rust, since it is currently available only for the AWS SDK for Rust.
  • Any move toward IAM policy evaluation or multi-account support, two stated limits at launch.
  • How quickly agent-built fixes close issues filed on the GitHub repository.

Sources