AWS Built a Free Local Cloud Emulator, and No Human Wrote a Line of It
AWS released AWS Emulator on November 11, a free tool that runs more than 200 AWS services on your own machine. Autonomous AI agents wrote every line of its code, with no human review.

A Local Copy of AWS, Free
Amazon Web Services released AWS Emulator on November 11, a free tool that simulates more than 200 AWS services on your own machine. Developers point the AWS CLI or an SDK at a local endpoint and run their code as usual, without calling the real cloud and without paying for it. It installs as a Linux package or runs in Docker on Windows and macOS, and it needs no AWS account for local testing.
The tool covers 438 API models, which is a large share of the AWS surface. It is built in Rust, and it uses AWS API models as machine-readable specifications to reproduce service behavior, including expected responses and error messages. You can create S3 buckets, DynamoDB tables, and other resources locally, run your tests against them, and iterate without waiting on the network or watching the bill.
The practical problem it solves is real. AI coding agents such as AWS Kiro and Claude Code run many iterative test loops, and each loop against the real cloud costs time and money. A local emulator makes those loops fast and free, which is exactly what agent-driven development needs.
The Part That Made Headlines
The tool is useful. How it was built is the story. AWS says thousands of autonomous AI agents wrote every line of AWS Emulator's code, and no engineer on the team wrote or reviewed any of it. The source code is published on GitHub under a CC0 public-domain dedication, along with the raw commit history, so anyone can see what the agents did.
The project began as a weekend experiment by Matt Goldberg, a senior principal engineer at AWS. He gave the Kiro coding agent a three-step prompt: write tests from the AWS CLI, iterate until they pass against real AWS, then build an in-memory version that passes the same tests. Kiro came back after 15 minutes with a passable first version of Amazon Simple Queue Service.
From there, Goldberg and his team organized dozens of autonomous agents running on Amazon Bedrock AgentCore into three separate teams. The first team generates thousands of end-to-end fidelity test suites from API models and verifies them against the live AWS cloud. The second team finds gaps and issues. The third team implements fixes. Human engineers designed the system around specifications, repeatable evaluation, automated controls, and strict merge gates. Agent changes passed automated build, security, coding-standard, regression, and fidelity checks before merging.
AWS measures the emulator's accuracy through fidelity tests comparing API responses, HTTP status codes, error messages, and resulting resource states with actual AWS services. The company has published its Emulator Evaluation Suite on GitHub, so developers can see the tests used to measure compatibility.
What It Cannot Do Yet
AWS is unusually explicit about the gaps, and the honesty is refreshing. The emulator currently matches production services most closely for control-plane operations: creating, describing, updating, and deleting resources. Behavior outside the published tests remains unverified, and the GitHub repository carries a public backlog of known discrepancies under a section literally called things-that-do-not-work.
It does not evaluate Identity and Access Management policies or resource-based permissions the same way production services do. It does not simulate throttling, outages, or regional differences. It does not sandbox the code it runs. Each emulator instance operates as a single simulated AWS account. AWS says plainly that it is not for production or security testing.
That means the emulator is a development tool, not a replacement for the cloud. Use it to iterate fast and catch obvious mistakes locally. Then test against real AWS before you ship, because the emulator will not catch the failures that only happen in production.
Why This Matters
The emulator is interesting twice over. As a tool, it removes a real bottleneck for developers and for the AI agents that increasingly write their code. As a demonstration, it shows what autonomous agent teams can build when the system around them is designed well: specifications as ground truth, automated evaluation as the gate, and humans setting the rules instead of writing the code.
Whether you trust code no human reviewed is a fair question, and AWS's answer is that the evaluation suite is the review. The fidelity tests compare the emulator against live AWS, and the results are public. That is a defensible position, and it is also a testable one. Anyone can run the suite and check the claims.
If you build on AWS, try it this week. Point your CLI at localhost, run your test suite, and see what breaks. The AWS sovereign cloud test and the Genesis Mission cloud credits show how much AWS is investing in the platform around the emulator, and the VS Code agent update shows how the tools around it are changing too. The cloud is becoming something you can run on your laptop, and that changes who can build on it.


