AWS adds long-running bots to Kiro Crew agentic workspace

AWS is expanding Kiro, its AI-oriented development environment, with Kiro Crew bots, named long-running software agents that carry out tasks inside a shared workspace. The bots have their own memory and sessions, and AWS says they bill together through a developer's existing Kiro plan. The change comes as Kiro continues to shift from an IDE toward an agentic development environment.

What Kiro is now

Kiro is not the most popular code editor; Microsoft's Visual Studio Code holds that position in Stack Overflow's 2026 Developer Survey. Kiro might rank 27th among listed IDEs, but it now styles itself 'an agentic development environment' after a brief stint as 'an agentic IDE.' It is also described as 'an AI-powered development environment' that helps build software from prototype to production, and it is available as a desktop IDE, command-line interface, web app, mobile app, and personal AI agent called Crew.

Kiro is also the inheritor of Amazon Q Developer coding assistant users, who have been directed to migrate to Kiro. Kiro Crew is described as the persistent, open source development workspace, and it is more agent infrastructure that supports Kiro Web, IDE, and CLI.

How Kiro Crew bots work

'A bot is a new way to structure and drive agentic work,' a Kiro blog post says. 'Instead of working in isolated sessions, each bot is a named, long-running agent that you define and customize for a specific part of your work. It has its own memory and runs its own sessions so you stop tracking them one by one. Kiro Crew already learns from how you work. Bots change where that learning lives.'

AWS is launching the bots as part of an expansion of Kiro's remit, recalling similar efforts such as OpenAI dots and Grok Bot, as well as Claude Duet and Meta Muse. The bots can be given their own tools, permissions, and memories, which gives each agent a distinct identity. A possible downside is duplicated context if many agents process overlapping data without some form of shared cache.

Billing, memory, and example tasks

AWS says all Kiro bots bill together through a developer's existing Kiro plan. Each bot has its own memory and sessions, and learned context is intentionally kept separate so one bot's learning does not bleed into another's work. All bots operate within the same Kiro Crew workspace, use the same harness, and have access to the same codebase, project files, MCP, skills, and tools, so they do not need to independently rediscover project structure.

Because AWS Kiro subscriptions are sold as credits that convert to tokens at different rates depending on the underlying model, figuring out how Kiro Crew agents spend tokens may require analytics. As a thought experiment, AWS suggests an intake bot for reading trouble or job tickets, labeling them, flagging duplicates, and drafting reproduction steps for bugs. Other examples include a build-health bot for checking the CI pipeline periodically and surfacing test failures, a code review bot for reviewing pull requests and other delegated responsibilities, and a release-readiness bot for confirming merged changes have tests and seeking release sign-off.

Sources: The Register