Amazon has released Kiro Crew, an open-source framework for deploying multiple AI coding agents that work autonomously across sessions and tasks, after the tool was internally adopted by more than 39,000 developers at the company. Originally built inside Amazon under the name MeshClaw, the system enables developers to hand off asynchronous coding assignments to AI agents that continue working without active oversight, handling jobs like incident analysis, ticket sorting, codebase migrations, and pull request monitoring. The platform is now available under the Apache 2.0 license and can run locally or on developer-controlled infrastructure.
The workspace provides persistent, multi-session development capabilities through shared memory, reusable skills, scheduled tasks, concurrent agents, and custom applications. Agents retain project context, execute tasks during off-hours, hand off work to subagents, and connect with external tools via the Model Context Protocol and webhooks. According to the announcement, the project attracted 500 contributors within six months without a company mandate. Community members testing the platform report that Crew can burn through tokens considerably faster than the standard Kiro command-line interface, raising operational cost questions.
Bolin Chen, senior software engineer at Amazon, Zejiang (Joe) Guo, principal software development engineer at AWS, and Zezhen Xu, senior software engineer at Amazon, write that the trio "wanted something simple that wasn't available internally, a way to kick off a task, walk away, come back to something worth reviewing, and to run several tasks at once instead of babysitting one prompt at a time." The authors add that Kiro Crew "ships 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." The system orchestrates agents through the Agent Client Protocol, offering live visibility into agent activity with a view that displays each agent's plan, tool invocations, approval checkpoints, and outcomes as they unfold.
The internal traction figures have drawn particular notice from industry observers. Muhammad Ishaq, CTO at Nextbridge, comments that the internal uptake numbers "say more than the feature list honestly," noting that 39,000 users and 500 contributors in half a year without a mandate signals "people were actually solving real problems with it, not just trying a new tool because it existed." Mathi M, senior solution architect at Verizon, highlights that "having background subagents handle parallel jobs without blocking the main workflow is exact target engineering builders need," and notes that "stacking context across sessions instead of starting from scratch will save tons of redundant prompt engineering." The platform runs atop the Kiro CLI and supports existing Kiro configurations including steering files, skills, and custom agents, and ships with integrations for Slack, Telegram, and WeCom for macOS, Linux, and Windows. According to the release, the project welcomes community involvement in both development and roadmap decisions, with Kiro and AWS engineers serving as initial maintainers and external contributors able to join as the ecosystem grows. For organizations already experimenting with CI/CD migrations, dependency management triage, scheduled automation, and long-running workflows, the open release offers a framework designed around real-world use cases proven at scale inside one of the world's largest engineering organizations. The organic growth inside Amazon suggests the tool carved out a niche before formal promotion, a pattern that typically indicates genuine workflow fit rather than top-down adoption pressure.