Kiro is launching a new generation of AI bots. They are designed to perform ongoing software development tasks with less human intervention. The bots are now generally available in Kiro Crew. They can coordinate multiple AI agents, maintain project knowledge, and perform recurring tasks autonomously.
Unlike conventional AI coding assistants that operate in individual sessions, Kiro bots are designed to take responsibility for entire workflows. Each bot has its own memory, workspace, and permissions. This enables the bot to manage tasks over extended periods without requiring developers to repeatedly provide context or instructions.
According to Kiro’s announcement, the bots can initiate multiple agent sessions and monitor their progress. They will also intervene if tasks deviate from their original intent. If a decision requires human input, the bot can request approval while continuing with other tasks.
The new functionality is designed to save developers time when coordinating AI agents. Instead of reviewing individual sessions, users can assign responsibilities to specialized bots and monitor their activities via dashboards.
For example, one bot can monitor continuous integration pipelines for build errors, while another bot checks pull requests against established coding standards. A third bot could manage release preparations, review documentation, and assess test results before requesting final approval.
Bots can also follow schedules. Developers can configure a bot to check for errors in the code every 30 minutes. The bot can also prepare release notes every Friday.
Separation of privileges and memory
Kiro emphasizes that each bot operates within its own security boundaries. A research bot may have permissions to read files and search the internet, without modifying code. A development bot can independently run scripts and make changes, as long as they fall within its assigned permissions.
Each bot maintains its own memory. This prevents knowledge about separate workflows from becoming intermingled. Developers can inspect, modify, or delete stored information. They can also adjust permissions as circumstances require.
Users can choose between AI models for individual bots and their agents, for example, to reduce costs by assigning simple tasks to less expensive models.
The launch signals a further shift within AI-powered software development: from tools that respond to individual commands to systems capable of managing continuous processes.