AWS Launches Kiro Crew for 39,000-User Open-Source AI Engineering Workflows

3 articles · Updated · InfoWorld · Aug 4

Kiro Crew lets enterprises run persistent, multi-agent software workflows across repositories, tickets and pull requests, with scheduling, memory, approval gates and audit logs aimed at work that continues without a developer present.
AWS said the platform can be self-hosted on laptops, containers or virtual machines without an AWS account, a design analysts said could give CIOs more control over security, compliance and shadow AI-agent use.
More than 39,000 Amazon builders adopted the internal precursor MeshClaw in under six months, and AWS is pairing the release with reference apps for worktree management, issue triage and long-running task execution.
Analysts said the open-source release could cut release-cycle delays and reduce prompt babysitting, but warned enterprises still need policies for access, approvals, memory retention and cost control before deploying persistent agent swarms.
Open standards such as ACP and MCP underpin Kiro Crew, yet it relies at launch on the proprietary, credit-metered Kiro CLI, potentially forcing teams using Claude Code, Codex or Devin to build their own connectors.