AWS releases Emulator code for agents to test more than 200 services without the cloud

- Reading
- 5 min
AWS Emulator runs locally, simulates the public APIs of more than 200 AWS services, and was written entirely by coding agents, according to InfoQ.
What the AWS Emulator Is
AWS has open-sourced the code of the AWS Emulator, a program that runs locally and exposes an HTTPS endpoint compatible with the public APIs of more than 200 AWS services, according to a report by InfoQ published on November 16, 2026. Developers and coding agents can create resources, receive the same errors AWS would return, and iterate without provisioning anything in the cloud or paying for the emulated operations.
The approach addresses a common friction point for those automating infrastructure with AI: testing a flow against real AWS costs money, and if the agent makes a mistake, it can create unwanted resources or get stuck in a production account.
How to Connect Without Changing Code
To use it, simply point the AWS CLI or an AWS SDK to the emulator using environment variables such as AWS_ENDPOINT_URL; the application code doesn't change. On Linux, it installs as an RPM or Debian package for x86_64 or aarch64. On macOS and Windows, AWS documents a short Dockerfile, but doesn't publish a ready-made image.
Teams working with Rust get an embedded mode: the emulator runs inside the test process itself and connects its HTTP client to the AWS SDK for Rust. There is also a skill called aws-emulator that starts the emulator for coding agents and checks the project against the fidelity tests published with each release.
Written by Agents, for Agents
The emulator is written in Rust and generated from more than 400 Smithy API models that AWS publishes for its SDKs, according to Swaminathan Sivasubramanian. AWS states that agents wrote and tested the emulator's own code.
Matthew Goldberg, senior principal engineer at AWS, describes the process as three teams of agents working in sequence:
- One team writes fidelity tests from the public API models and verifies them against real AWS
- Another reproduces those tests against the emulator and opens issues when it finds a divergence
- A third implements each fix that's flagged
Every change goes through a merge gate that checks compilation, code standards, security, and fidelity regressions. This entire pipeline runs on Amazon Bedrock AgentCore Runtime Instances, AWS CodePipeline, and AWS CodeCommit.
Uneven Fidelity Across Services
How closely it matches real AWS behavior varies by service. AWS publishes versioned fidelity tests and per-service scores with each release; a test only passes when every call matches AWS, including status codes, side effects, and the exact errors. Unimplemented operations return HTTP 501 with the error code NotImplemented.
According to AWS, fidelity is higher in control-plane APIs, while data-plane behavior and support for CloudFormation, Terraform, and AWS CDK vary from service to service. The documented limitations include:
- No evaluation of IAM or resource-based policies: the emulator isn't meant for testing security
- A single account and region shared per process
- In-memory state, which disappears when the process stops
- No testing for long-running use or load
AWS's own evaluation suite includes tests that the emulator still doesn't pass.
How It Compares to LocalStack, Floci, and moto
Developers already have local alternatives for simulating AWS, such as LocalStack, Floci, and moto. The AWS Emulator sets itself apart by being generated from AWS's own API models, publishing official fidelity scores with each release, and being dedicated to the public domain, but it doesn't replace those tools in every flow, especially permission testing and infrastructure-as-code pipelines.
What the Community Is Already Testing
Developer Josh Builds published his own breakdown of the fidelity numbers shortly after launch:
AWS just released a free local emulator of its APIs, made for coding agents and written entirely by agents. I counted its published fidelity tests: S3, DynamoDB and SQS pass about 2 in 3. CloudFormation passes under 1 in 5.
Josh Builds, developer
Yurii Serhiichuk compared the emulator with existing alternatives and already filled a packaging gap that AWS left open:
AWS shipped AWS Emulator today - think LocalStack/Floci but native. Rust, built by agents (so hopefully it gets updated pretty fast). No public docker image though, and no GitHub action. So I filled that part.
Yurii Serhiichuk, developer
What Changes for Those Integrating Agents with AWS Infrastructure
In short: the AWS Emulator gives those building pipelines with coding agents a disposable, no-cost environment to validate API calls before touching a real AWS account. This matters especially for those already using Amazon Q Developer, Claude Code, or another coding agent integrated with infrastructure: the loop of error, fix, and retry that used to run against the production cloud (with the risk of creating an improper resource or generating a charge) can now run entirely locally.
The important caveat is not to confuse API fidelity with tested security. Since the emulator doesn't evaluate IAM policies or resource-based policies, teams that rely on agents to validate permissions still need a real environment (or tools like LocalStack, which cover that case) before releasing any flow into production. The in-memory state, which disappears when the process ends, also makes it better suited to quick CI tests than to long debugging sessions.
For anyone evaluating the tool, the fidelity scores published with each release are the first place to look: they show, service by service, whether it's worth trusting the emulator for that specific flow or whether it's better to keep testing against real AWS. The code is available under the CC0 1.0 Universal public domain dedication, except for third-party code in the vendor/ directory, which is licensed under the Apache License 2.0, and the AWS Emulator is now available in general availability at no additional service cost.
Translated from the Brazilian Portuguese original · Read the original
Uber creates framework to decide whether an AI agent's action is appropriate in context
The company detailed a framework that goes beyond identity and API permissions to assess whether an autonomous agent's action makes sense in context, covering the agent's entire lifecycle.





