AWS open-sourced Kiro Crew, a persistent development workspace that lets AI coding agents run multi-step tasks unattended — across sessions, tools, and time.
The core pitch from AWS’s Darko Mesaros: “an application layer that turns AI coding agents into always-working, self-learning, autonomous teammates.” It’s orchestration, not another IDE. (InfoWorld)
What it does:
- Persistent context — remembers project state across sessions; you come back to progress, not a restart
- Unattended execution — multi-step tasks run without you watching; recurring jobs run on a schedule
- System heartbeats — monitors external systems until a condition is met, then triggers the next step
- Multi-channel access — desktop app, web dashboard, CLI, plus Slack and Discord to continue the same work thread
- Self-learning — adapts to how you work over time
- Runs on your hardware — local or remote, not locked to a cloud provider
Kiro Crew sits above individual coding agents. Where Cursor or Claude Code handles a single session, Kiro Crew coordinates work that spans hours or days — merging PRs, running recurring jobs, picking up where the last session left off. The session view shows an agent thread working through decisions like weighing PR title options, with a merged pull request as output.
NOTE
Kiro Crew is separate from Kiro IDE, AWS’s agentic code editor. Crew is the orchestration layer; the IDE is where you’d do hands-on editing. They’re complementary, not competing.
Available now as open source on GitHub and listed on Product Hunt. No pricing tier — it’s free and self-hosted.