AWS released AWS Emulator on November 11, 2026 — a free tool that simulates more than 200 AWS services on a developer’s own machine — and AWS is openly touting the strangest part: autonomous AI agents wrote the code, and no human engineer wrote or reviewed a single line of it. Why it matters: this is the first time a major cloud provider has shipped a broad developer tool built and maintained end-to-end by agents, and it is aimed squarely at the bottleneck those same agents created.
November 11, 2026
What happened
AWS Emulator replicates AWS service APIs locally and exposes a local endpoint that the AWS CLI, the AWS SDKs, and coding agents can use with the same APIs they use against the real cloud — developers redirect their tools by setting the AWS_ENDPOINT_URL environment variable, and existing code and tests run against the simulated services without modification and without an AWS account. It is built from more than 400 of AWS’s published API models covering more than 200 services, is written in Rust, runs on Linux, macOS, and Windows (including as a Docker container), and is free. AWS released the source under a CC0 public-domain dedication on GitHub, together with its full commit history.
Built — and maintained — by agents
“The emulator was written entirely by autonomous agents. Our team hasn’t written or reviewed any of the code,” wrote Matt Goldberg, the AWS senior principal engineer behind the project. The project began as a weekend experiment inspired by Anthropic’s earlier demonstration of Claude agents building a C compiler: Goldberg prompted the Kiro coding agent, and it produced a passable emulator of Amazon SQS in about 15 minutes; by the end of the weekend it had roughed out the rest of AWS, and his team was using it that Monday. The agents now run on Amazon Bedrock AgentCore in three teams — one writes “fidelity tests” (real scenarios like creating a queue, sending a message, deleting the queue) and checks them against real AWS, a second runs those tests against the emulator and files issues for failures, and a third fixes them, one agent per issue. Every change passes an automated gate where other agents check coding standards and security, and Goldberg says hundreds of agents now fix thousands of issues a day.
The limits are public, too
AWS and outside analysts are candid that a pass against the emulator is not a guarantee of behavior in the real cloud. IDC analyst Devin Pratt told InfoWorld the emulator does not yet simulate throttling or outages, snapshot and restore, or per-Region availability differences, and it does not sandbox the code it executes. Unite.AI’s coverage adds that the emulator does not evaluate IAM or resource-based policies the way AWS does, so a request AWS would deny could succeed locally — making it a build-and-test tool, not a security test environment. Unimplemented operations return an error rather than silently succeeding, each instance is scoped to one account and Region with in-memory resources deleted when it stops, and the known fidelity gaps are published in a GitHub folder literally named “things-that-do-not-work.” The obvious comparison is LocalStack, the existing AWS emulator, which this year put its free version behind an account requirement and limited it to noncommercial use.
Why it matters
Coding agents write code faster than teams can validate it, and validating against the real cloud is slow and billed by the call — so the agent era created a testing bottleneck that a local, free, agent-speed stand-in directly attacks. Just as notable is the production claim itself: AWS is not quietly using agents, it is marketing “no human wrote or reviewed this code” as a selling point, with the safety net moved into specs, fidelity tests against the real cloud, and automated merge gates. Whether that holds up in production is exactly what the public issue backlog will show over the coming months — and AWS says it will share more at its upcoming re:Invent conference.
Frequently asked questions
What is AWS Emulator?
A free, open-source (CC0) tool from AWS that simulates more than 200 AWS services locally, built from over 400 published AWS API models. The AWS CLI, SDKs, and coding agents connect to it through a local endpoint using the same APIs as the real cloud.
Did AI really write all of AWS Emulator’s code?
That is AWS’s own claim: senior principal engineer Matt Goldberg wrote that the emulator “was written entirely by autonomous agents” and that his team “hasn’t written or reviewed any of the code.” Agents organized in three teams also maintain it, testing against real AWS and fixing filed issues.
Can AWS Emulator replace testing in the real cloud?
No. It does not simulate throttling, outages, or regional availability differences, does not evaluate IAM policies the way AWS does, and does not sandbox executed code. AWS positions it for building and fast iteration; applications still need validation against real AWS before production.
Sources: GeekWire, InfoWorld, SiliconANGLE, Unite.AI (all Nov 11, 2026)