AWS Open-Sources Kiro Crew for Asynchronous AI Coding Agents
AWS open-sourced Kiro Crew, an agent-orchestration tool already used internally at Amazon by more than 39,000 developers, under an Apache 2.0 license.
Overview
AWS has open-sourced Kiro Crew, a tool that lets AI coding agents keep working across sessions on tasks like incident investigation, pull-request monitoring and multi-hour migrations without a developer actively supervising each step, according to Kiro’s announcement blog post. The project began as an internal Amazon side project called MeshClaw and had already been adopted by over 39,000 Amazon builders before its public release, according to Kiro and independently confirmed by InfoQ.
What We Know
- Kiro Crew was built by three Amazon engineers — Bolin Chen, Zejiang (Joe) Guo and Zezhen Xu — who wrote in the announcement that they “were inspired by the momentum of OpenClaw and various tools with self-learning agents taking over AI work, but we needed something that met our security requirements for internal development work,” according to Kiro.
- In under six months as an internal tool, Kiro Crew grew to nearly 500 internal contributors shipping 597 updates at an average pace of 143 weekly commits, according to Kiro.
- The tool runs on the Kiro CLI and orchestrates multiple agents through the Agent Client Protocol (ACP), with an Activity view that shows each agent’s plan, tool calls and results as they happen, according to Kiro and InfoQ.
- Security is built in from the start: Kiro described the tool as shipping with “defense in depth from day one, an OS-level sandbox, denied-by-default commands, suspicious-pattern blocking, input validation, sensitive-path blocking, credential redaction, and a signed audit log of every action,” according to Kiro.
- The project is released under an Apache 2.0 license and hosted at github.com/kirodotdev/KiroCrew, where it has already drawn more than 3,400 stars, according to the project’s GitHub repository.
- Setup guides in the repository cover macOS, Linux and Windows, along with connectors for Slack, Telegram and WeCom, according to Kiro.
- Governance runs through a steering committee of maintainers listed in the project’s MAINTAINERS.md file, with proposals filed and debated as pull requests, according to Kiro.
- The release follows AWS’s participation in the Agent Plugins 1.0 packaging standard earlier this month, which listed Kiro among its launch-day compatible clients alongside ChatGPT, Codex, Cursor, GitHub Copilot and VS Code, as previously reported.
What We Don’t Know
- Kiro has not disclosed how the 39,000-builder adoption figure breaks down by team or role inside Amazon, or how it compares with usage of other AI coding agents at the company.
- No independent security audit of the sandboxing and credential-redaction claims has been published; the security description comes from Kiro’s own announcement.
- Kiro has not said how external, non-Amazon adoption has trended since the public release, or whether it plans paid or hosted tiers alongside the open-source project.
Analysis
Kiro Crew’s release adds to a wave of AI coding tools moving from single-prompt assistants toward persistent, multi-session agent orchestration — letting a developer hand off a task like a repository migration or an on-call investigation and return later to reviewable progress rather than a stalled prompt. AWS pairing that shift with an unusually detailed security list — sandboxing, command denylists and a signed audit log — reflects the same tension other coding-agent vendors have faced this year: broader autonomy for agents means broader access to code, credentials and CI systems, which raises the stakes if that access isn’t tightly constrained.