AWS has launched AWS Emulator, a local copy of its cloud APIs that runs on a developer’s laptop, and says the software is written by AI agents. Its launch post says it “is built and maintained by agents from the AWS API models”, and that it “has no additional service fee”.
The public repository shows the pace. It held 29,684 commits when we read it on 11 November, from an “Initial emulator scaffold” on 21 October to 10 November, about 1,480 a day by our arithmetic. The 100 newest were all by two accounts named IssueFixerAgent and IssueFinderAgent.
The AWS CLI, the AWS SDKs and coding agents talk to the emulator “over a local HTTPS endpoint using the same AWS APIs they would use in the cloud”. It is written in Rust, installs from RPM or Debian packages on Linux and runs in Docker on macOS and Windows.
AWS pitches it at agent workflows. The post says “A coding agent can create resources, check the result, and iterate again without waiting on a deployment in a Region”, and names Kiro and Claude Code as examples of the coding agents it has in mind.
Share of single-service fidelity tests that AWS Emulator passed in release v1.0.322, for selected services, from the score file AWS attaches to each release. The counts and shares are our arithmetic.
Business Pill · The Model and the Harness
An evergreen Business Engineer explainer on the model and the harness. It teaches the general idea only and does not discuss AWS Emulator.
The key insight: As we read it, AWS has turned its own API models into a specification that agents code against, and its fidelity tests into the scoreboard. The public repository logged 29,684 commits in about 20 days, and by our count the 11 November release passed 24,243 of 42,937 tests, about 56%.
How the Agents Build It
AWS describes two groups of agents. “One group of agents looks for gaps”, comparing the emulator with real AWS and with the published Smithy models, and files an issue for each. “Another group picks up an issue, writes tests, implements the behavior, and reviews the change for correctness, performance, and security”.
Each change runs through AWS CodePipeline and the fidelity tests before it reaches GitHub, and “an independent security agent reviews changes adversarially”. The release notes for v1.0.322, published on 11 November, list 215 fixes, and the README says the repository “does not take PRs”.
Fifty-Six Percent, by Our Count
AWS publishes a score file with every release. A test passes only if the emulator “matches AWS on the response to every call”, per the post. By our count of the v1.0.322 file, 24,243 of 42,937 fidelity tests passed, about 56%, with 18,614 failing and 80 skipped.
The spread is wide. Among single-service tests, by our arithmetic, Amazon SQS passed 90% and Amazon S3 79%, AWS Lambda 46%, Amazon EC2 41% and AWS CloudFormation 16%. The user guide calls a score “a snapshot” and says failing tests are known gaps it publishes on purpose.
As we read the score file, the familiar storage and messaging services sit well above the average, while services with long multi-step workflows sit well below it: SQS at 90% and S3 at 79%, against CloudFormation at 16%, by our arithmetic on single-service tests.
As we read the post, the emulator is aimed at the loop coding agents run: AWS says a coding agent can create resources, check the result, and iterate again without a deployment, so a local endpoint shortens each pass of that loop.
AWS News Blog, 11 November 2026
“Because the code is written by agents, an independent security agent reviews changes adversarially, both while the fix is being made and again in the release pipeline as a final check before anything ships.”
Three Implications
FOR DEVELOPERSAWS says the emulator has no additional service fee and works with the AWS CLI and SDKs once environment variables point them at the local endpoint.
FOR SECURITYAWS says not to use the emulator to test whether an application is secure, because it does not evaluate IAM policies and resource-based policies the way AWS does.
FOR AWSAWS says agents keep the emulator current with AWS as the API models change, and its user guide says the emulator gets updates many times a day.
The framework argues: “The surface, not the model, governs autonomy.”
As we read AWS’s post, the emulator is part of that surface for cloud work: a local endpoint and a public test suite that a coding agent can run against. The framework predates the emulator and does not discuss it.
What Is Not Established
AWS does not say in the post or guide pages we read how many agents work on the emulator or what running them costs. Our commit figures cover the repository’s public history as we read it; work done before the 21 October scaffold commit, if any, is not in that history.
Our pass rates count each test in the score file once and compare services on single-service tests; AWS reports its own k of n scores per service, and the guide says to compare scores only on the same version of the fidelity tests. The emulator “doesn’t evaluate IAM policies the way AWS does”. We have not run it.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
AWS Emulator is a local copy of AWS’s APIs with no additional service fee, and AWS says agents build and maintain it. Its public repository logged 29,684 commits in about 20 days, and by our count it passed 24,243 of 42,937 fidelity tests, about 56%, in its 11 November release.
A note on sourcing. We read AWS’s 11 November launch post in full and the AWS Emulator user guide’s overview and fidelity pages. The commit, author and release figures come from the public aws/aws-emulator repository, read through GitHub’s API on 11 November. The pass counts are our own tally of the score file attached to release v1.0.322, counting each test once. We did not contact AWS for this piece.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
