Boost Team Productivity with New Team Member Kiro Crew

Kiro Crew is an open-source personal AI agent workspace that runs continuously on local or remote machines. Started as an Amazon internal side project MeshClaw, it was adopted by over 39,000 developers within six months and is now publicly operated under the Apache 2.0 license. Unlike simple AI assistants, Kiro Crew functions as an 'autonomous agent' that works on its own at scheduled times and remembers past context.

The architecture of Kiro Crew consists of a user interface, a gateway, and an agent backend. The gateway is a single asynchronous process implemented in Python and aiohttp, coordinating requests from various channels such as the dashboard, CLI, and Slack into a single session. Actual inference and tool execution are handled by the agent backend, Kiro CLI, which works without additional configuration by leveraging existing .kiro settings.

Key features include independent session management and Autopilot mode. Autopilot automates tasks step-by-step through a plan-approve-execute flow, requesting user judgment upon failure to prevent infinite loops. Additionally, the memory system stores preferences, projects, history, semantic memory, etc., divided between files and a DB (SQLite + FAISS), allowing new sessions to continue from previous context.

The Schedule feature allows agent tasks to be automatically executed using cron expressions or time intervals. This enables tasks such as issue triage, preliminary research, and PR reviews to be performed even when no humans are present, while adhering to security policies and governance rules. A template gallery makes it easy to apply various presets for code quality, security, operations, and more.