Open-source AWS emulator built by AI agents

Amazon Web Services has introduced AWS Emulator, a free utility that allows developers to simulate AWS services locally on laptops, development servers, or within continuous integration (CI) pipelines. The most striking aspect of this release is its origin: the tool was written entirely by autonomous AI agents. AWS confirms that no human engineers wrote or reviewed the codebase. While the emulator itself carries no cost, users remain responsible for paying standard fees for actual AWS resources when they eventually deploy their applications.

Autonomous Development and Agent Roles

The creation of AWS Emulator relied on three distinct teams of AI agents, each assigned specific responsibilities. AWS notes that agents perform more effectively when tasked with well-defined objectives. This structured approach allowed the company to delegate the entire coding process to machines while maintaining oversight through predefined boundaries.

“The emulator was written entirely by autonomous agents. Our team hasn’t written or reviewed any of the code. We could do this because the guardrails we’ve put in place gave the agents space to work independently and us the trust that they don’t color outside these lines,” Matthew Goldberg, a senior principal engineer at AWS, explained.

The first agent team generates tests based on public AWS API specifications. These tests verify actions such as creating message queues, sending messages, and deleting queues against real AWS services. The resulting library contains tens of thousands of tests that can be replayed locally without contacting AWS infrastructure. The second team executes these tests against the emulator to identify behavioral discrepancies. It reports failures and clusters issues that share common underlying causes. The third team addresses these reported problems, with individual agents handling specific defects. Before any modification is merged into the main codebase, it must pass build checks and undergo review by other agents for coding standards, security compliance, and fidelity. Changes that fail these rigorous checks are returned for revision. Hundreds of agents operate simultaneously, resolving thousands of issues daily.

Implementation Methods and Use Cases

Developers can integrate AWS Emulator using four primary methods: an AI coding agent, a Linux package, Docker, or embedded mode. Each method has unique prerequisites. The Linux package operates as a local process, whereas Docker supports Windows, macOS, and Linux environments. Embedded mode runs the emulator within a Rust test process.

This flexibility enables developers to test code interactions with AWS services, including cross-service calls, directly on their own machines. Integration tests can run in CI pipelines without requiring an active AWS account or allocated resources. Additionally, developers can prototype applications locally using AI coding agents before moving to production deployment.

Limitations and Data Handling

AWS Emulator is strictly designed for development and testing, not for production workloads. It does not evaluate AWS Identity and Access Management (IAM) or resource-based policies, making it unsuitable for security testing. The documentation recommends using a single region and account for all requests to an emulator instance, as the tool does not isolate resources by region or account. Fidelity levels vary across different services and features. Behavior verification occurs against AWS in US West (Oregon), except for global services, which are checked in US East (N. Virginia).

All emulated resources reside in memory and are lost when the emulator stops. AWS has not validated the tool for long-running processes or load testing scenarios. By default, the emulator transmits anonymous usage and error data to AWS, though this reporting can be disabled. No requests, responses, or resource data are sent externally.

The source code and fidelity tests are publicly available on GitHub under a CC0 public-domain dedication. The repository organizes every issue filed by the agents into fixed and unresolved folders. AWS publishes per-service fidelity scores with each new release to provide transparency regarding the emulator’s accuracy.