AWS says autonomous agents built its AWS emulator

AWS Emulator is designed to let developers build and test AWS applications on a computer or in a CI environment.[1] The AWS CLI and software development kits can communicate with it through a local HTTPS endpoint, using the same AWS APIs as in the cloud.[2] AWS says emulator calls return in milliseconds, so developers and coding agents can test changes and fix errors locally.[1] The company says there is no additional service fee for using the emulator, although customers pay for AWS resources when they deploy to AWS.[1]

AWS says autonomous AI agents wrote the emulator’s code and that its team neither wrote nor reviewed it.[4] The company says the implementation uses Rust to keep CPU and memory use low on a laptop.[2] AWS says the emulator draws on more than 400 API models covering more than 200 AWS services.[3] Its development process includes separate teams managing a test suite, finding emulator issues, and fixing them.[4] AWS says hundreds of agents work concurrently on the codebase and have fixed thousands of issues a day.[4]

AWS says supported operations return the same results as AWS for valid calls and the same errors for invalid calls; unsupported operations return a 501 error.[1] New services and APIs appear in the emulator as AWS fills in and verifies their behavior release by release.[1] The company says a growing fidelity suite contains tens of thousands of tests across AWS services.[4] However, IDC analyst Devin Pratt told InfoWorld that the emulator does not yet simulate throttling, outages, or regional differences in service availability, and does not sandbox the code it runs.[3]

Linux users can install the emulator from RPM or Debian packages, while macOS and Windows users run it in Docker; AWS says it does not publish a ready-made Docker image for those users.[2] The source code is published under the CC0 1.0 Universal public-domain dedication.[1] AWS says updates from agents are released many times a day.[5] Its documentation says the emulator is for building and testing, not production workloads, and should not be used for security testing because it does not evaluate IAM or resource-based policies as AWS does.[5] AWS says it plans to share more about the project at re:Invent.[6]