AWS Emulator: 200 Cloud Services on Your Laptop, Written Entirely by AI Agents
AWS Emulator runs more than 200 AWS services locally for free, with no account needed. Amazon says autonomous AI agents wrote every line, and the code passed 56.6 percent of its fidelity tests.

AWS released AWS Emulator on November 11, a free, open-source program that pretends to be more than 200 AWS services on your own machine. The AWS Emulator is built from 438 API models, needs no AWS account, and charges no fee. The part everyone is talking about is who wrote it: Amazon says autonomous AI agents wrote every line, and the repository had about 30,681 commits by November 13, all signed by automated identities.
The emulator runs locally, so a developer can build and test software that talks to AWS services without sending anything to Amazon's cloud or paying for cloud usage while testing. That is the practical pitch, and it is a good one. Cloud bills during development are a real cost, and a local emulator removes them entirely for the services it covers.
What the AWS Emulator actually does
At launch it passed 56.6 percent of 42,857 published fidelity tests, or 24,243 of them. Fidelity here means how closely the emulator behaves like the real AWS service, and 56.6 percent is honest rather than impressive. It is enough to be useful for development and testing, and not enough to trust blindly. The public repository accumulated about 29,684 commits in its first three weeks and about 30,681 by early November 13, which is a pace no human team sustains.
The build story is the headline. Three teams of agents working on Bedrock AgentCore built the emulator, with humans setting the tests and merge gates, the checks that code must pass before it is accepted into the project. Senior principal engineer Matthew Goldberg wrote on About Amazon that the emulator was written entirely by autonomous agents and that his team has not written or reviewed any of the code. That is a striking claim from a company whose cloud runs a large share of the internet.
The release was announced on November 11, with coverage and a new release, v1.0.338, continuing through November 12 and 13. The speed of follow-up releases is part of the story too: agents that write code can also iterate on it fast.
What it cannot do
AWS is explicit about the limits, and they matter. The emulator does not evaluate IAM or resource policies the way AWS does, and it never denies a request signed with its own key. That means it cannot tell you whether your permissions are set up safely. If your code works against the emulator, it might still fail, or worse, be insecure, against real AWS.
It should not be used for production workloads or security testing. That is not fine print. It is the difference between a development tool and a safety tool, and confusing the two is how breaches happen. Use the emulator to build and test logic. Use real AWS, with real policies, to verify security.
Why this release matters beyond AWS
The emulator is interesting on its own, but the bigger story is what it proves about AI-written software. Until now, the most visible AI coding output has been demos, side projects, and autocomplete. This is infrastructure software for the world's largest cloud provider, written by agents, with humans setting the tests and the gates instead of writing the code.
That changes the job of the human engineer. Goldberg's team did not write or review the code, but they did define what correct means, through 42,857 fidelity tests, and they control what gets merged. The work moves up a level: from writing code to specifying behavior and judging results. That is a real shift, and it is the same shift visible across the industry, from AWS rebuilding its free tier to Step Functions workflows becoming tools that AI agents can call.
The honest number in this story is 56.6 percent. It is the fidelity score, and it is also a measure of how far agents have come and how far they still have to go. A tool that is right a little more than half the time is useful for development and dangerous for anything else. AWS knows that, which is why the limits are stated so plainly. The emulator is free, it is open source, and it is worth trying. Just do not mistake it for the real thing.


