Amazon just released a free copy of its own cloud that runs on your laptop. The twist: AWS says no human wrote or reviewed a single line of code.
The tool is called AWS Emulator, and thousands of autonomous AI agents built it. Here is how it works, what it can and cannot do, and whether you should trust it.
What Exactly Happened?
On 11 November 2026, AWS Senior Principal Engineer Matthew Goldberg published a post on About Amazon titled How we built the agents that built AWS Emulator. It says the emulator was written entirely by autonomous agents, and that the team has not written or reviewed any of the code.
What AWS Emulator is
- A local emulator that answers AWS API calls on your own machine, a dev server or a CI pipeline.
- Free to use. You only pay for real AWS resources when you deploy to AWS.
- Released under the CC0 public domain license. The source is on GitHub at aws/aws-emulator.
- Reportedly built in Rust and generated from AWS’s public API models. The starting point was a server that handles requests for 438 AWS API models.
- Coverage is reported as more than 200 services. One developer write-up counts 426, so numbers differ by how services are counted.
- Works with existing AWS SDKs, the command line and AI coding agents.
How the agents built it
AWS describes three teams of agents, each with a narrow job:
- Test writers build fidelity tests from the AWS API models and check them against real AWS.
- Issue finders replay those tests against the emulator and file failures.
- Fixers each take one issue and send a change through a merge gate that checks the build, coding standards, security and fidelity.
AWS says tens of thousands of fidelity tests exist, and hundreds of agents work at once, fixing thousands of issues a day. The project began as a weekend experiment inspired by Anthropic’s demonstration that agents could build a C compiler. AWS admits the first prototype had gaps and that guardrails took substantial engineering. In one early build, agents even wrote their own Python interpreter for Lambda, so the team told them to reuse existing libraries.
Known limits
InfoQ reports that the docs say the emulator does not evaluate IAM permissions. Other coverage says it also skips cross-Region behavior. That matters, because permission bugs are common in real cloud apps. The repository tracks open and fixed issues in public folders.
Why It Matters
For Users
If you use apps built on AWS, nothing changes today. But the tool shows how fast software can now be produced and tested by agents, which may speed up the tools developers use behind the scenes.
For Businesses
Local testing means faster, cheaper development. Developers can test code without creating cloud resources, which cuts cost and risk. The key rule: use the emulator for quick feedback, then confirm on real AWS, especially for permissions and security.
For the AI Industry
This is one of the clearest examples yet of a large, production-style project written without human code review. It is also a test case for a hard question: how do you trust code nobody read? AWS’s answer is heavy automated testing against the real system. For more on coding agents, see our post on DeepSeek’s open-source coding agent, and our guide to agents vs chatbots vs assistants.
Social Media Reaction
We could not verify specific posts well enough to quote them, so here is the shape of the discussion in coverage.
Supporters are impressed that public fidelity tests and a public commit history let anyone check the work. They like a free, public domain tool that coding agents can use in a safe test loop.
Skeptics say practitioners are split on trusting an emulator nobody reviewed. They note the IAM gap and the risk that an emulator passes tests that real AWS would fail. Some also point out that a company marketing agents is the same one reporting on its own results.
What Should You Do Next?
- Try it on a non-critical project. Point your AWS CLI or SDK at the emulator for unit and integration tests.
- Keep a real AWS check. Run final tests on a real account, especially for IAM, networking and cross-Region behavior.
- Read the issue lists. Check the repository’s “things that now work” and “things that do not work” folders before you depend on a service.
FAQ
Is AWS Emulator free?
Yes. It is free and released under CC0. You only pay for real AWS resources when you deploy.
Did humans really write none of the code?
That is what AWS says: the emulator was written entirely by autonomous agents, and the team neither wrote nor reviewed the code.
Does it replace testing on real AWS?
No. It does not evaluate IAM, so you should still test on real AWS before launch.
How many services does it cover?
Coverage is reported as more than 200 services, built from 438 AWS API models. Counts differ by source.
Conclusion
AWS Emulator is useful on its own, and it is a milestone for agent-written software. The takeaway for builders: fast local testing is great, but trust still comes from checking against the real thing.
Also read: Adobe MAX 2026 and the MCP interface for outside agents. Follow AI Trends Setter for daily AI news.
