AWS Releases Local Cloud Emulator Built Entirely by Autonomous AI Agents

Amazon Web Services has introduced AWS Emulator, a headless local application that mimics more than 200 AWS services without requiring an active cloud account or service fees. Developed to give developers and coding agents a fast, free testing loop, the tool accepts commands from standard AWS CLI tools and software development kits by redirecting API endpoints locally, leaving application code untouched.

Autonomous Agents at Scale

What sets the release apart is its creation method. The project originated from a single prompt to an agent named Kiro, which drafted an initial Amazon SQS emulator in roughly 15 minutes. AWS subsequently divided development across three specialized teams of autonomous agents: test writers generating scenarios from public API models, issue finders replaying tests to log failures, and fixers submitting patches.

According to Matthew Goldberg, the project utilized hundreds of concurrent agents addressing thousands of issues daily. The automated release pipeline operated using AWS Developer Tools with CodeCommit as the primary repository before publishing the code to GitHub.

Performance and Fidelity

The initial v1.0.322 release builds on 438 AWS API models, running in Linux environments or Docker containers on macOS and Windows. Fidelity scores vary significantly across components. While services like AWS KMS and Amazon SQS achieved high scores of 93.9% and 89.8% respectively, AWS CloudFormation lagged substantially at 15.8%. Overall, the emulator successfully passed roughly 56.6% of its designated fidelity tests.

"AWS Emulator gives a fast, free test loop against AWS-shaped APIs." — Source Material

What to watch next: Monitor future updates to see if agentic workflows can improve low-scoring areas like CloudFormation and cross-service workflows.