RU

AWS Emulator runs AWS API locally: tests without the cloud or an account

Published: 2026-11-12 · Author: AI Release · @ai_release1
AWS Emulator runs AWS API locally: tests without the cloud or an account

⚡ The gist in 5 seconds

🔍 What happened

The aws/aws-emulator repository on GitHub now hosts AWS Emulator — a local AWS API emulator. The repository lives in the official aws organization, the README lists 29,684 commits, and the Hacker News post scored 721 points and 222 comments. The idea is simple: your code talks to the emulator on your own machine, dev server, or CI instead of real AWS. No account is needed, and there's no separate fee for the emulator — you only pay for AWS resources when you actually deploy to the cloud.

There's no graphical interface. In server mode, the emulator serves HTTPS on 127.0.0.1:9297 with a self-signed certificate; you then set environment variables, point AWS CLI or any SDK at that endpoint, and run your code as usual. On Linux, an RPM or Debian package is installed and launched with the aws-emulator command; on Windows and macOS, you use a container built from a short Dockerfile. There's an embedded mode for Rust: the emulator is added as a dev-dependency and runs inside the test process, with one isolated instance per test — no URL, certificate, or environment variables; the emulator's HTTP client plugs directly into the SDK client. A separate path is prepared for AI agents: aws agent-toolkit add-skill --skill-name aws-emulator --region us-east-1, requiring AWS CLI 2.35.0 or newer.

Building from source is a heavy exercise. One crate per AWS service, at least 100 GB for target/, Rust with Cargo and a C compiler like GCC or clang, and a multi-core machine. Linux x86_64 and aarch64 are supported; on Windows and macOS, only Docker. The commands: cargo build -p aws-emulator-router, then cargo run -p aws-emulator-router.

🧠 Our take

The main point here isn't the binary itself but who released it. Market experience shows the local AWS emulation niche has long been occupied by third-party projects, and the question for them has always been the same: how faithfully do they replicate the behavior of the real cloud. Now the vendor itself is offering the reference implementation — and promises to publish fidelity scores with every release on the releases page. Our assessment: this is a strong argument for teams tired of chasing discrepancies between "works on the emulator" and "works in production."

What changes in practice is CI economics. If tests don't hit the cloud, the pipeline sheds the bill for test resources, the hassle of keys, and the wait for an environment to spin up. But accuracy matters more than speed: an emulator that doesn't respond like AWS saves money and frays nerves. The README should be read literally — the gaps are honestly laid out in the issues/things-that-do-not-work/ and issues/things-that-now-work/ folders, which is rare: usually such lists live in a wiki and rot.

Who wins? We believe AWS first of all: the cheaper testing is, the more teams take integration all the way to deployment — and they pay AWS anyway. The losers are third-party emulators, which now face a reference implementation from the platform's owner. But the cost of entry stings too: 100 GB and a multi-core Rust build in embedded mode is not "install and forget."

⚠️ What we don't know

The most important point in the README sends you to the releases page for the numbers, and no specific fidelity scores appear in the text — meaning we can't tell from this fragment how faithfully, say, S3 or DynamoDB are emulated. There's no service list, no performance data, and no comparison with real AWS. The 721 points on Hacker News signal hype, not project maturity. And separately: the README does not permit security testing on the emulator, so it's not a tool for security tests.

💬 What's worth debating

  • Move CI to the official emulator while it doesn't cover all services, or is it cheaper to keep a separate AWS test account and not worry about discrepancies?
  • Fidelity scores are computed by the vendor itself: is that an honest measure of AWS parity or a marketing number that needs to be verified with your own tests?
  • 100 GB of disk and a multi-core machine for embedded mode — a reasonable price for test isolation, or a barrier that will scare off small teams?

Prediction: within a quarter, the service list in the README will grow, and at least two releases will ship with updated fidelity scores — this can be checked on the releases page without speculation.

Source: GitHub (aws/aws-emulator)

🤖 AI summary
  • Date: 2026-11-12
  • Source: github.com
  • Tags: AWS, DevOps, CI/CD, cloud, testing, MachineLearning, AI
  • Key facts:
    • AWS Emulator spins up a local AWS API endpoint: requests go to the emulator , not to the cloud, and no AWS account is needed.
    • Linux — an RPM or Debian package and the aws-emulator command; Windows and macOS — a container built from a Dockerfile; Rust tests get an embedded mode.
    • For building and testing only: the README explicitly forbids production workloads and security checks on the emulator.
    • The aws/aws-emulator repository on GitHub now hosts AWS Emulator — a local AWS API emulator.
    • The repository lives in the official aws organization, the README lists 29,684 commits, and the Hacker News post scored 721 points and 222 comments.
    • The idea is simple: your code talks to the emulator on your own machine, dev server, or CI instead of real AWS.
    • No account is needed, and there's no separate fee for the emulator — you only pay for AWS resources when you actually deploy to the cloud.
    • There's no graphical interface.
#AWS#DevOps#CI/CD#облако#тестирование#MachineLearning#AI
← PreviousProvenance vs. Fabrication: How to Keep a Made-Up Value Out of a Real CallNext →Samsung bets 144 trillion won on AI chip packaging

Source: github.com · post in Telegram