AWS Emulator and the engineer's job when agents write the code

Amazon Web Services has launched AWS Emulator, a free, locally hosted software suite that replicates hundreds of AWS cloud services on a developer's workstation, letting teams validate code without connecting to the real cloud. SiliconANGLE reported that the entire service was designed and written exclusively by AI agents, with not even a single line of its code written or reviewed by a human engineer. The tool is available to download and is open-sourced under a Creative Commons CC0 public-domain dedication license.

The problem it addresses comes from the speed of AI coding agents such as AWS Kiro and Claude Code. SiliconANGLE reported that these agents generate code faster than teams can verify it, and testing against real AWS infrastructure has become nearly impossible because of latency and cost. The AI Economy noted that developers testing against the real cloud often needed a dedicated account costing $10,000 or more. Existing third-party emulators cover only a small fraction of AWS' service catalog and drift out of sync as services evolve, according to SiliconANGLE.

The project began as an experiment. AWS Senior Principal Engineer Matt Goldberg, inspired by Anthropic's report that a group of AI agents built a working C compiler, asked the Kiro coding agent to generate tests from AWS' command-line tools, adjust them until they passed against the real cloud, and then build a stand-in that passed the same tests without it. SiliconANGLE reported that Kiro generated a functional replica of Amazon Simple Queue Service within 15 minutes. Goldberg then spent the weekend having Kiro work through the rest of AWS, and his team began testing their own software against the result.

AWS organized dozens of autonomous agents running on Amazon Bedrock AgentCore into three teams, The AI Economy reported. The first builds an answer key: agents study Smithy files, the blueprints AWS uses to describe each service, and devise traces recording exactly how AWS responds to a series of actions. Every trace is confirmed against the live cloud before being added to a library that now holds tens of thousands of them. The second team replays those traces against the emulator to hunt for problems, grouping shared root causes into issues filed under "issues/things-that-do-not-work." The third team claims one issue at a time and works on the fix.

Before code enters the GitHub repository, it must clear a merge gate, The AI Economy reported. Code that does not run or fails its own tests is rejected, and reviewer agents examine each change for coding standards, security and accuracy. Tom "Spot" Callaway, a principal open-source technical strategist at AWS and a contributor to the Emulator, described those reviewers as asking, "Is this good code? Is this the right way to do it? Is this an optimized way?" A fix that fails any check goes back to the agent for another try.

The emulator is written in Rust and runs as a local HTTP server, accessed through a command-line interface on Linux, Windows or macOS, or as a Docker container. Developers redirect their AWS CLI or SDK calls to localhost for rapid, offline validation. The AI Economy reported that it supports more than 400 AWS services whose API definitions are publicly available on GitHub, though Callaway acknowledged it does not yet handle every scenario within each one, and AWS plans to keep expanding coverage.

Callaway stressed that the tool contains no AI itself. "None. It's not AI," he said, explaining that there are no large language models inside it and that it is "just a bunch of Rust code." He also drew a line at its limits: "Is it going to be performant like a real-world service? No. Is it going to get all of the cases right? No." The emulator is not designed to replace AWS, and neither Callaway nor AWS provided minimum computer specifications.