기본 콘텐츠로 건너뛰기

AWS Emulator Tutorial: Let Coding Agents Test AWS Locally

In short: AWS Emulator is a free, open-source program from AWS that answers AWS API calls on your own machine, so your code and your coding agent can create S3 buckets, DynamoDB tables and SQS queues without touching a real account or bill. I installed release 1.0.322 on an Ubuntu box, pointed an unchanged boto3 script at it with the environment variables from the official guide, and it worked on the first try. The catch is fidelity: by my own count of the published test scores, only about 56% of its fidelity checks pass overall, so treat it as a fast test bed, not a copy of AWS.

What AWS Emulator is (and what it is not)

AWS released the emulator on GitHub as aws/aws-emulator, licensed CC0 1.0 (the vendored framework code is Apache 2.0). It is a single local process with no graphical interface. It serves one HTTPS endpoint, https://127.0.0.1:9297 by default, and every supported service is mounted on that one endpoint. The AWS CLI and the AWS SDKs reach it through standard environment variables, so your application code does not change. There is no service fee; you only pay when you later deploy to real AWS.

The launch essay on Werner Vogels' blog frames it as a test environment for the era when agents write much of the code: an agent can run the same create, read and fail loops it would run against AWS, but locally and in seconds. The official guide lists three use cases: local testing, CI pipelines with no AWS account or resources, and prototyping with AI coding agents.

What it is not: a production runtime, a security test tool, or a multi-region simulation. The guide is explicit that it does not evaluate IAM and resource policies the way AWS does, keeps everything in memory (all resources vanish when it stops), does not separate resources by account or region, and is verified against AWS only in us-west-2 (us-east-1 for global services).

Step 1: Install it

Linux (native package)

AWS publishes RPM and Debian packages for x86_64 and arm64. On Debian or Ubuntu, the guide downloads aws-emulator_amd64.deb from downloads.aws.com and installs it with sudo apt-get install -y ./aws-emulator.deb; on Fedora or Amazon Linux 2023 it is a one-line sudo dnf install of the RPM URL. The package puts the program in /opt/aws-emulator/ and the aws-emulator command in /usr/bin/.

Worth doing first: each package has a .sig file next to it. I imported the AWS publishing key, checked its fingerprint (98BC F336 A3BE 53B7 34E9 67C5 2E5B F49B D940 F624, matching the guide) and ran gpg --verify, which reported a good signature made on Nov 11, 2026. The package itself is about 171 MB and unpacks to a single binary of roughly 970 MB, so plan disk space accordingly. Then aws-emulator -v printed version 1.0.322, release date 2026.11.10.

macOS and Windows (Docker)

There is no prebuilt image. The guide gives a short Dockerfile based on amazonlinux:2023-minimal that installs the RPM and starts the emulator with -b 0.0.0.0 inside the container. Two details from the guide matter: publish the port as -p 127.0.0.1:9297:9297 (the short form -p 9297:9297 exposes it to your whole network), and mount ~/.aws/emulator so your SDK can read the certificate and config the container writes. The sample command also caps memory at 2 GiB; if the container goes over, it exits with code 137 and the emulated resources are lost.

With a coding agent

If you work in an agent, the README offers an aws-emulator skill, installed with aws agent-toolkit add-skill --skill-name aws-emulator --region us-east-1 (AWS CLI 2.35.0 or later). The skill installs, starts and connects the emulator and can check which parts of your project it is expected to handle. I did not test the skill path in this run; my test below used the plain server mode.

Step 2: Start it and point your SDK at it

Run aws-emulator and leave that terminal open. I started it with --telemetry off (usage telemetry is on by default; you can also set EMULATOR_TELEMETRY=off or add [telemetry] enabled = false to ~/.aws/emulator/config.toml) and --idle-timeout 900 so it would shut itself down after 15 idle minutes. On startup it wrote four files to ~/.aws/emulator/9297/: ca.pem, config, emulator_credentials and truststore.p12. Those local keys are generated fresh for each run and work only with that run of the emulator.

In a second shell, set the variables from the official guide:

export AWS_ENDPOINT_URL=https://127.0.0.1:9297
export AWS_SHARED_CREDENTIALS_FILE=~/.aws/emulator/9297/emulator_credentials
export AWS_CONFIG_FILE=~/.aws/emulator/9297/config
export AWS_CA_BUNDLE=~/.aws/emulator/9297/ca.pem
export AWS_EC2_METADATA_DISABLED=true
export AWS_DISABLE_HOST_PREFIX_INJECTION=true
export AWS_REGION=us-east-1
export AWS_DEFAULT_REGION=us-east-1

The guide also says to clear any other AWS credential environment variables in that shell, because they take precedence over the credentials file. Then check readiness with curl --cacert ~/.aws/emulator/9297/ca.pem https://127.0.0.1:9297/__admin/health, which returned ok for me. The quickest sanity check is sts get-caller-identity: on the emulator the account is always 000000000000 and the ARN is arn:aws:iam::000000000000:user/emulator. If you see any other account number, your calls are not going to the emulator.

Step 3: Run real code against it

I used boto3 1.44.1 and a short script with no endpoint or key in the code. Results from that run:

  • STS: returned account 000000000000, as documented.
  • S3: created a bucket, uploaded a small text object and read it back intact.
  • DynamoDB: creating a table named ab failed with the same ValidationException wording AWS uses for names shorter than 3 characters; a valid table accepted a put_item and returned it with get_item.
  • SQS: created a queue, sent and received a message. The returned queue URL still reads sqs.us-east-1.amazonaws.com/000000000000/jobs, but boto3 kept sending the calls to the local endpoint, so do not panic when you see an AWS hostname in the output.
  • Virtual-hosted S3: forcing a client to localhost with virtual-host addressing sent the request to lm24-demo-bucket2.localhost and failed with an SSL error, exactly the case the guide warns about.

One honest oddity: in my very first pass, a put_item issued immediately after create_table returned ResourceNotFoundException. On the second pass the table was already ACTIVE and the waiter returned at once. Real DynamoDB also requires an active table, so add ddb.get_waiter("table_exists").wait(...) after creating tables in tests anyway; it costs nothing here and keeps the same test valid against AWS.

Step 4: Check fidelity before you trust a green test

Every release ships a scores.json.gz on the GitHub releases page. I downloaded the one for 1.0.322 (scored Nov 11, 2026 UTC) and tallied it myself, counting each test unit toward every service it lists. That is my own arithmetic, not an official per-service score, but it gives a realistic picture:

  • Overall: 42,937 units, 24,243 passing (56.5%), 18,614 failing, 80 skipped.
  • Stronger: EventBridge 71.5%, Step Functions 70.0%, SNS 69.3%, DynamoDB 68.1%, Secrets Manager 68.0%, SQS 67.8%, S3 62.7%.
  • Weaker: CloudWatch Logs 57.3%, EC2 38.8%, Systems Manager 23.8%, CloudFormation 18.9%.

The release notes are long lists of fixes, and the guide says AI agents update the emulator many times a day, so these numbers will move. The official way to check your own project is to list the operations it calls, look them up in the manifest.jsonl of the aws-emulator-evaluation-suite repository, and compare with the scores for the release you run, or ask the agent skill to do it.

AWS Emulator vs LocalStack vs Moto

Tool Price How you run it Coverage and limits (from official pages)
AWS Emulator 1.0.322 No service fee; CC0 1.0 license Linux deb/rpm, Docker image you build, or embedded in Rust tests One endpoint for all services; in-memory only; no IAM policy evaluation; telemetry on by default; per-release fidelity scores
LocalStack for AWS Hobby free (non-commercial); Base $39 and Ultimate $89 per license per month, billed annually; Enterprise custom Local emulator; running tests in CI is included from the free Hobby plan 30+ services on Hobby, 55+ on Base, 110+ on Ultimate; state persistence and basic IAM enforcement from Base
Moto 5.2.5 Free; Apache 2.0 Python library (@mock_aws decorator or context manager) or moto_server standalone Best fit for Python unit tests; install per service with pip install 'moto[s3,...]' or moto[all]

Editor's note

Good for: integration tests that call several AWS services, CI jobs that should never create real resources, and agent loops that need quick, disposable infrastructure. Restarting the process gives you a clean state for free.

Not for: testing IAM or bucket policies, multi-region behavior, load tests, long-running dev environments, or anything you would call production.

Gotchas I hit or verified:

  • InvalidClientTokenId: a leftover access key variable in the shell overrides the emulator's credentials file. With a fake key set, my STS call failed with exactly this error; clearing the variable fixed it.
  • SSL errors on S3: keep 127.0.0.1 in AWS_ENDPOINT_URL. With localhost, some SDKs (JavaScript, Java, Go by default) put the bucket in the hostname, and the emulator rejects that.
  • "Could not connect to data-127.0.0.1": host-prefix injection; set AWS_DISABLE_HOST_PREFIX_INJECTION=true only in the emulator shell, because the guide notes it breaks those operations on real AWS.
  • Emulator gone after the terminal closes: it runs in the foreground; in my first attempt a backgrounded process ended with its shell. Use a dedicated terminal, a service manager, or Docker.
  • Port 9297 in use: start with -p 9300 (or -p 0 outside containers) and change the port in every variable.
  • Big build for embedded mode: building from source needs at least 100 GB free for target/, per the README.

FAQ

Does AWS Emulator cost anything?

No. The README and guide say it has no additional service fee; you pay only for AWS resources once you deploy to AWS.

Can I use it with JavaScript, Java or Go, not just Python?

Yes. It works with the AWS CLI and the AWS SDKs through the same endpoint variable. Some SDKs need an extra setting to trust the emulator's certificate or to disable host prefixes; the "Pointing your AWS SDK" page has a per-SDK table.

Does it keep my data between runs?

No. Everything lives in memory and is deleted when the process stops, along with the files in ~/.aws/emulator/<port>/. That is a feature for tests: each start is a clean slate.

Related on lightmind24: Agent Skills: One SKILL.md for Claude Code, Cursor, Copilot · Claude Code Errors Fixed: Login, PATH, 429 and Usage Limits · Promptfoo Tutorial: Test Your LLM Prompts Before You Ship

Start the emulator, run your test suite once against it, then check the fidelity scores for the services your code depends on before you trust the green checkmarks.

Screenshots: see the official GitHub README and AWS documentation screenshots, plus my own terminal captures from this test run, at the end of this post. Images: GitHub (aws/aws-emulator), AWS documentation, and lightmind24 own test run.

Last verified: 2026-11-12

Sources: aws/aws-emulator README · AWS Emulator releases (v1.0.322, scores.json.gz) · What is AWS Emulator? · Getting started on Linux · Getting started with Docker · Pointing your AWS SDK at AWS Emulator · Checking whether your project works · All Things Distributed launch essay · LocalStack pricing · Moto getting started

lightmind24 is an independent review site, not affiliated with Amazon Web Services, LocalStack or Moto.

Uploaded Image Uploaded Image Uploaded Image Uploaded Image

댓글

이 블로그의 인기 게시물

브루(Vrew) 요금제 정리: 무료 200크레딧으로 어디까지 쓸 수 있나

요약: 영상 자막 편집기 브루(Vrew)는 2026년 4월 22일부터 기능별 사용량 대신 '통합 크레딧' 하나로 요금을 매겨요. 무료 플랜은 매달 200크레딧이 들어오는데, 음성 분석(자동 자막)만 쓰면 최대 200분 분량이라 짧은 영상 몇 편은 충분하지만, 영상 초반의 Vrew 마크를 지우려면 유료 플랜이 필요해요. 이 글은 브루 공식 가격 페이지와 고객센터 공지를 2026년 10월 23일에 다시 확인해, 크레딧 계산법과 플랜 고르는 기준을 정리했어요. 목차 브루 요금 구조: 통합 크레딧이란 1크레딧으로 할 수 있는 일 무료 플랜으로 되는 것과 안 되는 것 유료 플랜 4종 비교표 월 구독·연 구독·기간 이용권, 무엇이 이득일까 상황별 추천: 학생·직장인·유튜버 크레딧이 모자랄 때: 사용 기록 지우개 업그레이드·환불 전에 알아둘 규칙 운영자 메모, 자주 묻는 질문 1. 브루 요금 구조: 통합 크레딧이란 예전 브루는 플랜마다 음성 분석 몇 분, AI 목소리 몇 자, 번역 몇 자처럼 기능별 한도를 따로 줬어요. 그런데 보이저엑스(브루 운영사)가 실제 이용 데이터를 보니 한두 가지 기능만 집중해서 쓰는 사람이 훨씬 많았다고 해요. 그래서 2026년 4월 22일 업데이트부터는 플랜마다 정해진 크레딧을 받고, 어떤 기능에 쓸지는 내가 정하는 방식으로 바뀌었어요. 플랜별 매달 지급량은 Free 200 · Light 2,000 · Standard 10,000 · Business 12,000크레딧 이에요. 공식 공지에 따르면 무료 사용자는 매달 1일에, 유료 사용자는 매월 결제일에 충전되고, 남은 크레딧은 다음 달로 넘어가지 않아요. 바뀐 뒤 무엇이 좋아졌나 공지에 나온 예시가 이해하기 쉬워요. Standard 플랜에서 AI 목소리만 쓰면 예전에는 최대 50만 자였지만, 지금은 1만 크레딧을 모두 AI 목소리에 몰아 쓸 수 있어 최대 100만 자가 돼요. 반대로 여러 기능을 골고루 쓰는 사람은 크레딧이 기능끼리 ...

에이닷 전화 vs 익시오: 통화 녹음·AI 요약, 내 통신사·폰에 맞는 앱 고르기

요약: 통화 녹음과 AI 요약을 쓰고 싶다면 먼저 내 통신사와 휴대폰 종류부터 확인해야 해요. LG유플러스 고객은 무료 앱 익시오, SKT 고객은 에이닷 전화가 가장 기능이 많고, 다른 통신사의 아이폰 사용자는 iOS에 들어 있는 기본 통화 녹음·전사 기능이 현실적인 선택이에요. 이 글은 2026-10-30에 SKT·LG유플러스·Apple 공식 페이지와 도움말을 직접 확인해 세 가지를 비교했어요. 왜 '앱 기능'보다 '내 조건'을 먼저 봐야 할까요 통화 요약 앱 비교 글을 보면 기능 목록부터 나열하는 경우가 많아요. 그런데 실제로 써 보려는 사람에게 가장 먼저 걸리는 건 기능이 아니라 이용 조건 이에요. 익시오는 LG유플러스 고객 전용이고, 에이닷 전화는 누구나 설치할 수 있지만 통역콜·baro 같은 기능은 SKT 고객에게만 열려 있어요. 아이폰은 운영체제 구조상 앱이 통화를 마음대로 녹음하기 어려워서, 통신사 앱이 안 되면 iOS 기본 기능에 기대야 해요. 그래서 이 글은 "어떤 앱이 제일 좋은가"보다 "내 통신사·내 폰에서 무엇을 쓸 수 있나" 순서로 정리했어요. 기능 차이는 그다음이에요. 한눈에 비교: 에이닷 전화 vs 익시오 vs 아이폰 기본 통화 녹음 구분 에이닷 전화 (SKT) 익시오 ixi-O (LG유플러스) 아이폰 기본 통화 녹음 (Apple) 누가 쓸 수 있나 통신사 구분 없이 앱 설치 가능. 단, 통역콜·baro·그룹통화 발신 등은 SKT 고객 전용 U+ 모바일 고객. 타 통신사, 만 14세 미만, 선불요금제 고객은 제한 (공식 사이트 FAQ상 알뜰폰 중 KB리브엠은 예외) 지원 국가·언어의 iPhone 사용자. 1명과 통화할 때만 녹음 가능 지원 OS·기기 Android 10 이상, iOS 16 이상 iOS 17 이상 iPhone 12 이상 / Android...

타입캐스트 vs 클로바더빙: 수익 나는 채널에 쓸 때 무료·유료 조건 비교

요약 타입캐스트와 클로바더빙은 둘 다 글만 입력하면 한국어 AI 목소리를 만들어 주는 서비스지만, 무료로 쓸 수 있는 범위가 꽤 다릅니다. 클로바더빙 무료는 매달 15,000자까지 쓸 수 있는 대신 수익이 생기는 채널에는 쓸 수 없고, 타입캐스트 무료는 처음에 주는 약 5분 분량의 다운로드로 끝나는 체험판에 가깝습니다. 광고 수익이 붙은 유튜브나 쇼핑몰 영상에 넣을 목소리라면 처음부터 유료 플랜 조건을 보고 고르는 편이 안전합니다. 왜 이 두 서비스를 같이 비교하나 쇼츠 내레이션, 수업 자료 더빙, 상품 소개 영상처럼 "내 목소리 대신 AI 목소리"가 필요할 때 한국 사용자들이 가장 많이 찾는 곳이 네오사피엔스의 타입캐스트 와 네이버의 클로바더빙 입니다. 기능 소개만 보면 비슷해 보이지만, 실제로 문제가 되는 건 음질보다 "이 목소리를 내 채널에 써도 되나" 입니다. 그래서 이 글은 기능 나열보다 무료 사용 범위, 출처 표기 의무, 유료 플랜 가격과 다운로드 한도를 중심으로 정리했습니다. 모든 수치는 2026년 10월 25일에 두 서비스의 공식 요금 페이지와 고객센터 도움말에서 직접 다시 확인한 내용입니다. 한눈에 보는 비교표 항목 타입캐스트 무료 타입캐스트 베이직 클로바더빙 Free 클로바더빙 Standard 월 요금 0원 9,900원 (연간 결제 시 10% 할인) 0원 19,900원 (네이버 클라우드 플랫폼에서 결제) 무료·포함 사용량 음성 생성·재생 무제한, 다운로드는 최초 3,000크레딧(약 5분) 1회 지급 월 30,000크레딧(약 60분) 월 15,000자, 다운로드 월 20회, 프로젝트 누적 5개 (매월 1일 갱신) 월 30,000자, 다운로드 월 40회, 프로젝트 누적 20개 쓸 수 있...