AWS Emulator runs 200+ AWS services locally, built by AI agents

AWS has released a free emulator that runs a local copy of more than 200 of its cloud APIs, and according to MarkTechPost all of its code was written by autonomous AI agents. It matters for two separate reasons: developers get an offline stand-in for a large slice of AWS's API surface, and everyone else gets a rare public number for how much of that surface agent-written code actually reproduces.

What did AWS release?

MarkTechPost reports that AWS released AWS Emulator, a local emulator of AWS APIs assembled from 438 API models and covering more than 200 services. It is free, and it is published under CC0, a public-domain dedication. Two details stand out. The first is authorship: autonomous agents wrote all of the code. The second is that the launch came with a measurement attached, with the emulator passing 56.6% of 42,857 published fidelity tests.

The word 'published' in that test count is doing real work. A pass rate you can inspect is a different object from a vendor claim you have to take on trust, and it gives anyone evaluating the emulator a concrete baseline to argue with.

How much of AWS does it actually reproduce?

The two coverage figures describe different things. More than 200 is a count of services; 438 is a count of API models, the finer-grained unit an emulator has to implement, since a single service can expose more than one API definition. Neither number says how deep the emulation goes inside any one API.

The fidelity score is the part that answers that, though only in aggregate. Of 42,857 tests, 56.6% pass at launch, which means 43.4% do not. An aggregate across an entire surface cannot tell you which services or operations sit behind the failures, so the practical move is to check the specific calls your own code makes rather than infer anything from the headline.

Why does it matter that agents wrote all of it?

Agent authorship is the part of this release most likely to be quoted out of context, so it is worth being precise about what it does and does not establish. It establishes that a broad, mechanical, well-specified body of work, namely implementing API definitions to match a reference, is now within reach of autonomous coding agents at a scale a human team would find expensive. MarkTechPost reports that agents wrote all of the code; it does not report how much human review or orchestration sat around that, and the 43.4% of failing tests are consistent with output that needed iteration.

For teams running coding agents, the transferable lesson is about task shape rather than tooling. The work completed here has a crisp specification and an automatic pass or fail signal. Work with those two properties is where agents tend to do well; work without them still needs a human definition of done before an agent starts.

Can you use it in place of real AWS?

Nothing in the report suggests it is meant as a production substitute, and the numbers argue against treating it as one. A local emulator is useful for what local emulators are good at: running and debugging code without provisioning resources, exercising error paths offline, and keeping feedback loops in continuous integration fast.

An API-level fidelity suite measures request and response conformance. That is not the same as reproducing everything around the API. Service quotas, regional behaviour and the way real systems interact under load are separate questions, and a 56.6% aggregate pass rate is a reason to verify your own paths rather than assume them. The sensible posture is to develop and test locally, then confirm against the real service before shipping.

Why the CC0 licence matters

CC0 dedicates a work to the public domain, which in practice means no attribution requirement and no obligation to open-source anything you build with it or fold it into a larger work. That removes the usual friction around adopting a large body of code: you can vendor it into a repository, bake it into a container image, or ship it inside a product without a licence review about derivative works.

For a component whose whole purpose is to sit inside developer tooling and test harnesses, that is a meaningful choice. Coverage you are allowed to copy is coverage you will actually use.

How to explore this in Grafy

The most direct route is Grafy Code, the AI coding studio, where you can work against the emulator with an agent in the loop, installing dependencies and executing code rather than only describing it. The AI coding studio walkthrough shows how that loop runs.

If you would rather read before you install, Reader can take the AWS API documentation and the MarkTechPost report as PDFs and answer questions across them, showing which pages were actually read. And if the broader question is how to judge what coding agents can build, our earlier coverage of Frontier3DBench approaches the same problem from the benchmarking side: agents producing artefacts that then have to be measured.

Frequently asked questions

What is AWS Emulator?

AWS Emulator is a local emulator of AWS APIs released by AWS and reported by MarkTechPost on 12 November 2026. It covers more than 200 AWS services, assembled from 438 API models, so software written against AWS can run against a stand-in on your own machine. Autonomous agents wrote all of its code, and it is published free of charge under CC0, a public-domain dedication.

Did autonomous agents write all of AWS Emulator's code?

Yes. MarkTechPost reports that autonomous agents wrote all of the emulator's code, and no hand-written portion is called out. That makes the release a claim about agent-written software as much as about AWS APIs. The open question is not whether the code was produced, but how much of it behaves like the services it imitates, which is what the 42,857 fidelity tests are there to measure.

What does the 56.6% fidelity pass rate mean?

It means that of the 42,857 published fidelity tests, 56.6% pass against the emulator at launch, so 43.4% do not. Because the figure is a single aggregate across more than 200 services, it does not identify which APIs or operations fail. If you plan to rely on the emulator, test the specific calls your application makes rather than reasoning from the headline number.

Source: MarkTechPost

---

Source: MarkTechPost — AWS Releases AWS Emulator: A Local Copy of 200+ AWS Services Built by AI Agents