AWS Launches Emulator to Scale Agent-Driven Development
The new tool provides coding agents with a local AWS environment, shifting the engineering focus toward architectural fidelity as AI accelerates code production.
AWS has released **AWS Emulator**, a local program that provides a simulated environment of AWS APIs for software testing without requiring an AWS account. According to reporting from *All Things Distributed*, the tool is designed to give AI coding agents a high-fidelity test environment, as agents now generate code faster than human teams can verify it in production.
Writing for *All Things Distributed*, **Swaminathan Sivasubramanian** notes that the emulator's development was driven by a hypothesis from **Matt Goldberg**, a Senior Principal Engineer on the AWS Developer Tools team. Goldberg proposed that agents could autonomously build working versions of services if provided with the real service and a written description—specifically the public API models written in **Smithy**, AWS's descriptive format.
This shift mirrors early distributed systems work at Amazon. Sivasubramanian recalls collaborating with **Werner** (Amazon CTO) and engineer **Al Vermeulen** to build "Alf," a distributed lock service based on **Leslie Lamport's** Paxos protocol. Because Paxos-based services are difficult to test, the team simulated Alf's environment within a JVM to inject faults and identify concurrency bugs before deployment. While a small team could previously build these simulations by hand, the current scale of AWS—which features over 200 services—makes manual environment creation impractical.
As detailed in the *AWS Emulator README* on GitHub, the tool allows developers to point the AWS CLI or SDK at a local endpoint. It supports a "server mode" via containers or packages and an "embedded mode" for Rust tests. The emulator is intended strictly for building and testing; AWS explicitly warns against using it for production workloads or security testing.