Kiro Crew launched August 4, 2026, and it’s a different product from the Kiro IDE you might already know. Where Kiro IDE is an AI code editor that plans before it codes, Kiro Crew is a persistent, autonomous coding workspace that keeps agents running 24/7 across sessions.
Think of it as the difference between a coding assistant and a coding teammate that never sleeps.
What is Kiro Crew?
Kiro Crew is an open-source agent orchestration platform built on top of the Kiro CLI. It started as an internal Amazon project called “MeshClaw” and was adopted by over 39,000 Amazon builders in less than 6 months. It now has 500+ contributors and 597 updates.
The core idea: hand off work to a crew of AI agents, walk away, come back to results. Not a single prompt-response. A coordinated team working on your behalf across sessions, tools, and time.
Key specs
| Spec | Value |
|---|---|
| Developer | AWS (Kiro team) |
| Launch date | August 4, 2026 |
| License | Open source |
| Origin | Internal Amazon project (MeshClaw) |
| Internal adoption | 39,000+ Amazon builders |
| Contributors | 500+ |
| Platform | macOS, Linux, Windows |
| Surfaces | Desktop app, web dashboard, TUI |
| Integrations | Slack, Telegram, Discord, WeCom |
How it works
Kiro Crew orchestrates agents through Agent Client Protocol (ACP). Every step is observable live in the Activity view: planning, spawning sub-agents, tool calls, approvals, and synthesis.
Persistent agents: Unlike a chat session that ends when you close the window, Kiro Crew agents persist across sessions. They carry memory, context, and progress.
Scheduled jobs: Run recurring tasks on schedules you define. Morning digests, PR monitoring, flaky test detection, deployment health checks. Jobs that don’t need reasoning run as plain scripts without a model call.
Webhook triggers: External events (Slack messages, GitHub webhooks, CI failures) can start agent work automatically. No manual kickoff needed.
Checkpointed tasks: Long-running work moves through checkpoints, validation, and retries while you work on something else or sleep.
Multi-agent coordination: Run several conversations concurrently with isolated context. Delegate independent work to subagents that return results to the parent conversation.
Self-learning agents
Kiro Crew agents are designed to improve over time:
Memory: Carries preferences, active-project context, and relevant history into new sessions. Does not start cold.
Lessons: Corrections become durable lessons that change later behavior. Workspace-scoped lessons for project-specific guidance.
Skills: Repeated patterns become reusable skills you can inspect, edit, or remove. Skills stay visible so you decide what your crew carries forward.
Security
Kiro Crew ships with defense-in-depth from day one:
- OS-level sandbox
- Denied-by-default commands
- Suspicious-pattern blocking
- Input validation
- Sensitive-path blocking
- Credential redaction
- Signed audit log of every action
Every step is observable live. You can watch how agents plan, spawn sub-agents, decide which tools to call, gate actions for approval, and synthesize results.
Apps: purpose-built interfaces
Some work doesn’t belong in a chat window. Kiro Crew introduces Apps: sharable, purpose-built interfaces that combine custom UI with agents, skills, schedules, integrations, and backend services.
Launch apps include:
- DevFleets: Work tree management
- Task Runner: Executing long-running tasks
- Issue Radar: Triaging issues and pull requests
- LaunchDarkly: Feature flag management directly in the workspace
You can create your own Apps with the App SDK.
How it differs from Kiro IDE
| Feature | Kiro IDE | Kiro Crew |
|---|---|---|
| Type | AI code editor | Persistent agent workspace |
| Session model | Per-session | Cross-session, 24/7 |
| Triggers | Manual | Cron, webhooks, manual |
| Multi-agent | No | Yes (coordinated crew) |
| Apps | No | Yes (purpose-built UIs) |
| Use case | Interactive coding | Long-running autonomous work |
If you use Kiro IDE, your configuration carries over to Crew. .kiro files, steering files, skills, and custom agents work without extra setup.
When to use Kiro Crew
- You need agents that work while you sleep
- You run recurring tasks (PR reviews, deployment checks, morning digests)
- You want webhook-triggered automation
- You need multi-agent coordination for complex projects
- You prefer open-source with full observability
When to use something else
- Interactive coding sessions: Kiro IDE, Cursor, or Claude Code
- Simple automation: GitHub Actions or cron jobs
- Self-hosted on your own infra: Hermes Agent (see below)
- Non-AWS ecosystem: CrewAI, AutoGen, or Hermes Agent
My take
Kiro Crew is the most serious entry in the “persistent agent workspace” category. The internal Amazon adoption (39,000 builders in 6 months) is genuine evidence that this solves real problems. The security posture is enterprise-grade. The Apps concept is interesting.
The main limitation: it’s tightly coupled to the Kiro ecosystem. If you’re not using Kiro IDE or CLI, the adoption path is steeper. And the AWS backing means the project’s direction is ultimately driven by Amazon’s priorities.
For teams already in the Kiro ecosystem, Crew is a natural extension. For everyone else, it’s worth evaluating against Hermes Agent and CrewAI.
FAQ
Is Kiro Crew free?
Yes, it’s open source. Kiro Crew runs on the Kiro CLI, which has a free tier. The Crew itself has no additional cost beyond the model API calls your agents make.
How is Kiro Crew different from Kiro IDE?
Kiro IDE is an AI code editor for interactive coding. Kiro Crew is a persistent agent workspace for autonomous, long-running work. IDE is for sitting at your desk. Crew is for work that happens when you’re not there.
Can I use Kiro Crew without Kiro IDE?
Yes. Kiro Crew runs on the Kiro CLI independently. You don’t need the IDE to use Crew. But if you use both, your configuration carries over.
Does Kiro Crew work with non-AWS models?
Kiro Crew runs on the Kiro CLI, which supports multiple model providers. Check the Kiro docs for current provider support.
Is Kiro Crew the same as MeshClaw?
MeshClaw was the internal Amazon project name. Kiro Crew is the open-source release. Same codebase, different name.
Related Articles
- I Used Kiro for a Week
- Hermes Agent Complete Guide
- Kiro Crew vs Hermes Agent
- Claude Code vs Aider vs Continue.dev
- Best AI Coding Tools 2026