AWS Emulator, launched November 11, 2026, runs 200+ AWS service APIs on your laptop or in CI for no fee. Agents wrote all of it. What works, what doesn't, vs LocalStack.
On November 11, 2026, AWS announced AWS Emulator ↗, a local stand-in for AWS that runs on your own machine. The AWS CLI, the AWS SDKs and coding agents talk to it over a local HTTPS endpoint with the same API calls they would send to the cloud. The @awsdevelopers ↗ account called it "generally available today."
The facts from the AWS News Blog post by Daniel Abib and the GitHub README ↗:
No additional service fee. You pay only when you deploy to real AWS. No AWS account is needed to run it.
Install: RPM or Debian packages on Linux. On macOS and Windows you build a Docker image from a short Dockerfile in the docs; AWS does not publish a ready-made image.
Endpoint: HTTPS on 127.0.0.1:9297, reachable only from your own computer by default. You point tools at it with AWS_ENDPOINT_URL plus the credentials and config files it writes to ~/.aws/emulator/9297/.
Keys: new keys on each start, IDs beginning with EMULATOR; requests signed with other keys are rejected.
Embedded mode for the AWS SDK for Rust gives each test its own in-process state.
Written in Rust and built from the public AWS API models. AWS's launch posts say that covers more than 200 services from more than 400 API models.
License: CC0 1.0, except the vendor/ folder (Apache 2.0). The README says the repository "does not take PRs"; bugs and fidelity gaps go to GitHub Issues.
Telemetry is on by default. Turn it off with aws-emulator --telemetry off or EMULATOR_TELEMETRY=off.
GeekWire ↗ and InfoWorld ↗ covered it the same day.
AWS says agents wrote every line. Matt Goldberg, the senior principal engineer behind it, wrote on About Amazon ↗: "The emulator was written entirely by autonomous agents. Our team hasn't written or reviewed any of the code."
His post describes three agent teams on Amazon Bedrock AgentCore. One writes fidelity tests, real scenarios such as creating an SQS queue, sending a message and deleting the queue, and checks each one against real AWS. One runs those tests against the emulator and files an issue for each failure. The third fixes one issue per agent. Every change passes a merge gate where other agents check compilation, coding standards, security and fidelity. Goldberg says hundreds of agents fix thousands of issues a day.
Per Swami Sivasubramanian's guest post ↗ on Werner Vogels's blog, it began with one Kiro prompt that returned "a passable first version of an SQS emulator in about 15 minutes."
AWS lists the limits itself. The emulator does not evaluate IAM policies or resource policies the way AWS does, so "a request that AWS denies might succeed in the emulator." There is no cross-account or cross-Region behavior. An operation it has not implemented returns HTTP 501, so the test fails instead of passing quietly. How closely it matches AWS "varies by service and feature." Each release publishes per-service fidelity scores (passing tests out of total), the tests are public in aws-emulator-evaluation-suite ↗, and known gaps sit in issues/things-that-do-not-work/.
Analysts quoted by InfoWorld add more. Devin Pratt of IDC said that "at launch, the emulator does not simulate throttling or outages, snapshot and restore, or per-Region availability," and that it does not sandbox the code it runs. Broadcom SRE Advait Patel warned: "If your tests never see a 429 or a timeout, your retry and backoff logic never gets tested."
"they decided to vibe code localstack, i see" — @moby763canary21 ↗
"I'd rip LocalStack out of CI this week." — @soycronus ↗
"agents built the sandbox that other agents will test against. this is either perfect or beautifully circular." — @abdullah0094 ↗
"Not a replacement, but a third testing layer." — Kento Ikeda, an AWS Community Builder, after running it next to the AWS Blocks mocks in CI (@ikenyal ↗)
"This is a real software factory" — Marc Brooker ↗
Reddit was not reachable for this story, so there are no Reddit quotes.
| Option | Cost to start | What it is | Main limit |
|---|---|---|---|
| AWS Emulator | No fee, no account | Local AWS API endpoint, CLI and SDKs | No IAM checks; fidelity varies |
| LocalStack | Free tier needs an account | Long-running AWS emulator | Free use is noncommercial |
| Moto | Free library | Mocks of AWS services | Mocks, not an emulator |
LocalStack's change comes from InfoQ ↗ (February 2026): the open-source Community Edition was folded into one image that needs registration, with paid plans from $39 per license per month billed annually. GeekWire adds that the free version is limited to noncommercial use. We have not tested either emulator for this post.
Try it if a coding agent writes your AWS code. The agent can create a DynamoDB table, write, read and retry against your laptop instead of a shared dev account, with no leftover buckets or surprise bills. Our coding tools guide and Claude Code review cover which agent to pair it with.
Keep a real-AWS stage before production. No IAM evaluation means a passing local test says nothing about permissions. No throttling means your retry code is untested. Check the fidelity score for each service you use before you trust a green run.
Keep real credentials out of the test shell. AWS's guide unsets them, and the code under test is not sandboxed.
LocalStack users on the free tier have the clearest reason to look. AWS's tool needs no account, costs nothing and is public domain. Paid LocalStack users should compare service by service first.
The bigger story is the method: a spec, a reference to test against and a strict merge gate let AWS ship code no one on the team read. If you review agent-written code, start with our AI code review tools roundup.