AI Agents Wrote AWS's Free Cloud Emulator

Amazon Web Services on Wednesday, Nov. 11, released AWS Emulator, a free tool that mimics more than 200 AWS services on a developer's own machine. AWS says autonomous AI agents wrote and tested all of its code, and that no engineer on the team wrote or reviewed any of it.

Key takeaways

• AWS Emulator runs offline, needs no AWS account and is built from more than 400 public API models covering 200-plus services.
• Three teams of AI agents write tests, find gaps and fix code; hundreds of agents now close thousands of issues a day, according to AWS.
• The code is public on GitHub under a CC0 public-domain dedication, along with a running list of behaviors that do not yet match real AWS.

What AWS Emulator does

According to the project's GitHub README, developers point the AWS CLI or an SDK at a local endpoint and run their code as usual, without calling the real cloud.

It installs as a Linux package or runs in Docker on Windows and macOS. The target user is as much the coding agent as the human: agents check their work by running code again and again, and doing that against the live cloud adds delay and cost.

How AI agents built the emulator

The project began as a weekend experiment by Matt Goldberg, a senior principal engineer at AWS. In a launch post, he said he gave the Kiro coding agent a three-step prompt: write tests from the AWS CLI, iterate until they pass against real AWS, then build an in-memory version that passes the same tests. About 15 minutes later he had a passable first version of Amazon SQS.

By the end of that weekend, Goldberg wrote, Kiro had built a rough version of the rest of AWS, and his team began using it the next Monday.

The production system splits work among three agent teams. One writes "fidelity tests," real-world scenarios checked against live AWS, now numbering in the tens of thousands. A second replays them and files issues for failures. A third assigns one agent per issue. Every change must pass a merge gate where other agents review standards, security and fidelity.

"Our team hasn't written or reviewed any of the code," Goldberg wrote.

Context: specs and tests as guardrails

Swami Sivasubramanian, AWS vice president of agentic AI and emerging tech, wrote in a guest post on CTO Werner Vogels' blog that he spent four nights supervising Kiro as it tried to rebuild a replication engine that once took about four months to write. With a spec and a test environment, it took about two hours.

"When agents write the code and check their own work, the engineer's job is to give them room to work on their own and to build the specs, tests, and merge gates that decide when that work is correct," Sivasubramanian wrote.

The question of how far to trust agent-written code is a live one; this site has covered research showing LLM bug-fixers broke working code more often than benchmarks suggested.

Limits and what's next

The emulator is not a full stand-in. IDC analyst Devin Pratt told InfoWorld, as reported by GeekWire, that it does not yet simulate throttling, outages or regional differences, and does not sandbox the code it runs. Known gaps sit in a public "things-that-do-not-work" folder.

It arrives after LocalStack, the best-known rival, this year limited its free version to noncommercial use behind an account requirement, GeekWire reported. AWS says it will share more at re:Invent.

FAQ

Is AWS Emulator free?
Yes. It has no service fee, needs no AWS account and is released under CC0.

Did humans review the code?
AWS says no. Engineers built the agent system and merge gate; agents wrote and reviewed the code.

Can it replace testing on real AWS?
No. AWS says it is not for production or security testing.

References

• How we built the agents that built AWS Emulator — Amazon News, Matt Goldberg (Nov. 11, 2026)
• AWS Emulator and the engineer's job when agents write the code — All Things Distributed, Swami Sivasubramanian (Nov. 11, 2026)
• aws/aws-emulator: AWS Emulator is a local emulator of the AWS APIs — GitHub
• AWS gives developers a free local stand-in for its cloud, built from scratch by AI agents — GeekWire (Nov. 11, 2026)