AWS releases a free local version of its cloud, built and maintained by AI agents

AI & Compute

AWS Emulator Goes Free: 200-Service Local Cloud Built by AI Agents

AWS Emulator is generally available free, covering 200-plus services in Rust, with agents on Bedrock AgentCore writing all code — no engineer review, per GeekWire.

By
Nathan Brooks
Filed
Channel
AI & Compute
Read
3 min read

Amazon Web Services on Wednesday made AWS Emulator generally available at no service fee, an open-source, Rust-written stand-in for its cloud that runs on a developer's own machine and covers more than 200 services, according to GeekWire. The company says software agents built and maintain the entire codebase, and GeekWire reported that no AWS engineer reviewed the code.

The tool installs from RPM or Debian packages on Linux and runs in Docker on macOS and Windows, with source available on GitHub. The AWS CLI, the AWS SDKs and coding agents talk to it over a local HTTPS endpoint using the same APIs they would call in the cloud. A test that creates an Amazon DynamoDB table, writes an item and reads it back runs without a deployment.

The emulator supports every service that has an API model in the public AWS API models repository — the same models the SDKs are generated from — so requests and error messages have the same shape as on AWS, according to the launch post. AWS pitches the tool at developers working with coding agents such as its own Kiro and Anthropic's Claude Code, which can create resources, check results and iterate without waiting on a cloud Region.

Who wrote the code, and who checked it?

AWS says the emulator is built and maintained by agents running on Amazon Bedrock AgentCore, AWS CodePipeline and AWS CodeCommit. One group of agents hunts for gaps by comparing the emulator's responses against real AWS and its code against the published Smithy models, filing an issue for each discrepancy. A second group writes tests, implements the behavior and reviews each change for correctness, performance and security. An independent security agent then reviews changes adversarially before release.

GeekWire reported that AWS says agents wrote all of the code without review by its engineers. The claim is the quiet point of the launch: a cloud vendor trusting its own tooling to a pipeline with no human gate.

Fidelity is scored, not asserted. AWS publishes a per-service fidelity score with each release — 40 of 50, for example — where a test passes only if the emulator matches real AWS on every status code, response body and error message. The tests, including failing ones, are public on GitHub. An operation the emulator does not yet implement returns an HTTP 501 error, and AWS cautions that a score can drop as it adds tests. A scoreboard that can only fall is at least an honest one.

What are the limits, and who is the competition?

The emulator has firm boundaries. It does not evaluate IAM or resource policies the way AWS does, works as a single account and has no cross-account or cross-Region behavior. A request that AWS would deny can succeed locally, and AWS tells users to test security settings in the real cloud, not against the emulator.

On privacy, the emulator listens only on the local machine by default and generates fresh credentials prefixed EMULATOR on every start. It sends anonymous usage data unless the user turns that off.

The release lands in a market LocalStack has led for years, and the timing is not accidental. LocalStack folded its free Community edition into a single image that requires an account and an auth token as of March 2026, a move that pushed some developers toward free alternatives. AWS is now that alternative, with the API models repository behind it and a fidelity score attached to each release.

The competitive picture is straightforward: a vendor-supplied emulator cannot fully impersonate its own IAM engine, and it does not have to. Developers who want fast iteration loops for agent-driven workflows now have a zero-cost option backed by the company that defines the APIs — while security testing stays where AWS would clearly prefer it, in the paid cloud.

Source: mlq.ai

Share this article:

More from Nathan Brooks

Nathan Brooks

Show full bio

Senior reporter covering people, launches and day-to-day practice in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.

858 articles

Related articles

« Previous article