AWS Emulator brings the cloud testing loop to your laptop
AWS Emulator lets developers and coding agents test AWS APIs locally. Explore its capabilities, agent-built development process, limits and deployment context.
A coding agent can change your application in seconds. Testing its next S3 upload or DynamoDB write should not mean waiting for another cloud deployment. AWS Emulator brings that feedback loop onto your own computer.
Launched on November 11, it is AWS’s local implementation of its service APIs. Existing SDKs and CLI commands connect to a local endpoint instead of a live AWS account. There is no additional service fee, and the source carries a CC0 public-domain dedication.
What you can test locally
AWS’s walkthrough demonstrates a practical development loop:
| Capability | What it lets you do |
|---|---|
| S3 storage | Create a bucket, upload an object and list it |
| DynamoDB | Create a table, write an item and read it back |
| CLI and SDK integration | Use familiar API calls against a local HTTPS endpoint |
| Embedded Rust tests | Run isolated instances inside test processes without a shared server |
Download and startup: The Linux x86-64 RPM was about 167 MB on November 12 (166,735,193 bytes), not the installed binary size. A DevelopersIO test measured 0.257 seconds from
docker runto an HTTP 200 health response for v1.0.322. That was one aarch64 Linux run using Docker 29.9.3, a 2 GB memory limit and a self-built 1.62 GB image, with other containers running. It excludes image build time and measures emulator readiness, not each service’s startup or guaranteed performance.
The CLI setup sets AWS_ENDPOINT_URL to localhost and uses emulator-generated credentials and certificates, with real credentials removed. This redirects requests; it does not mirror your existing cloud infrastructure. You create emulated resources, and their in-memory state disappears when the emulator stops.
Built for coding agents, and by them
AWS also uses AI agents to develop the emulator: they identify gaps, write tests and implement changes. Published fidelity tests compare its behavior with real AWS before release. That makes the agent-built process inspectable, not automatically complete.
Unsupported operations return HTTP 501. IAM policy enforcement and cross-account or cross-Region behavior are absent. Passing locally is still a step toward testing on AWS, not proof your production setup is correct.
Local testing already has options. LocalStack offers persistence and IAM testing on paid tiers; its free Hobby plan is non-commercial. Testcontainers can manage LocalStack containers, while Moto provides Python mocks. AWS adds a first-party choice, not a new category.
With Lambda Web Functions simplifying deployment and AWS Emulator bringing tests closer to the code, the path from an idea to a working application gets shorter. Developers and coding agents can try a change, inspect the result and fix mistakes locally before testing permissions, integrations and production behavior on AWS.
Loading content…