1. News
  2. AI models & agents
  3. AWS · AI coding
AWS · AI coding2 min read

Amazon's coding tool Kiro now lets you run a team of named AI helpers

AWS is adding named, long-running AI bots to its Kiro coding tool. All of them bill to one plan, so costs are worth watching.

By the Dashline deskSource: The Register
Amazon's coding tool Kiro now lets you run a team of named AI helpers
Photo: The Register

Amazon Web Services is adding "bots" to Kiro, its AI coding tool. Each bot is a named helper that keeps working on one part of a developer's job over time. AWS says all of them bill together through a developer's existing Kiro plan. That makes it easy to add helpers, and just as easy to lose track of what they use.

What changed

Kiro started as an AI-focused code editor. It now describes itself as an "agentic development environment," meaning software that can carry out tasks on its own. It comes as a desktop app, a command-line tool, a web app, a mobile app and a personal AI agent called Crew. People who used Amazon's older Q Developer coding assistant have been told to move to Kiro.

How the bots work

A Kiro Crew bot is a named agent that you set up for one area of work. Each has its own tools, permissions and memory, and runs its own sessions. AWS says this means you no longer track sessions one by one. AWS gave examples of what a team might include:

  • An intake bot that reads tickets, labels them and flags duplicates
  • A build-health bot that checks automated tests and reports failures
  • A code review bot that reviews proposed changes
  • A release-readiness bot that confirms changes have tests and seeks sign-off

The cost question

Each bot keeps its own memory, so several bots could end up processing the same material. The Register raised the concern that this duplication could add cost. Kiro plans are sold as credits, and credits convert to tokens, the units AI models use to measure work, at different rates depending on the model. The Register also noted developer complaints about runaway costs in Kiro.

An AWS spokesperson said the bots' learned context is kept separate on purpose, so what one bot learns does not bleed into another's work. All bots share the same workspace, tools and project files, so they do not need to rediscover your project. AWS did not give figures on what a team of bots costs.

What to do

If your team already pays for Kiro or similar tools, add bots one at a time. Check how many credits each one uses before adding the next. The Register expects that tracking token spending across bots will need some form of analytics.

The takeaway

Before adding AI bots to a coding plan, add them one at a time and check how many credits each one uses.

More articles

You should really read these.

Next step

Let’s talk