Image: Kiro
Why it mattersA developer running many agent sessions tracks what each one did. Named bots carry context forward. The person only reviews results.
A developer who runs several agent sessions at once spends time tracking which one finished. They also track which one stalled. They track what each one decided. The person is the thing holding the work together. That cancels most of the speed the agents were meant to add.
Kiro is the agentic development tool from AWS. Kiro said on 5 November it is adding bots to its Crew product. A bot is a named agent. It keeps one ongoing chat with its owner. It has its own memory. It runs its own sessions. It carries a goal from one task to the next.
What a bot is
Each bot has a name, an avatar, and a workspace inside Kiro Crew. The owner sets a goal. The owner picks a model. The owner chooses which tools the bot may use. The bot then runs its own sessions, watches them, and brings results back into its one chat. Kiro says a bot can ask the owner for a decision on one task. It keeps working on parts that do not depend on that answer.
A bot can also be put on a schedule. The post names two examples. A build-health bot checks the CI pipeline every 30 minutes. It surfaces only new failures. A release-readiness bot confirms merged changes have tests and release notes. It does that before each release. Kiro says the work log records scheduled activity. A finding or a decision that needs a person goes to the chat.
Scoped tools and separate memory
Each bot gets only the tools and permissions its work needs. Kiro's post gives two examples. A research bot can search the web and read files. It cannot change them. A deployment bot can run deployment scripts. It asks for approval before each change. A bot can request more tools. Nothing changes until the owner approves.
The memory stays per bot. A deployment bot learns about the release process. A research bot learns about the market. The two memories do not mix. That matters for a team sharing one long chat with a coding agent. The model confuses one project with another in that setup.
Kiro says bots run wherever Kiro Crew runs. The post names a laptop, a Mac Mini, or an EC2 instance. A bot can run on a stronger model. Its worker agents can handle routine checks on a cheaper one.
Kiro Crew's post is by Kyle Seaman, a product lead, and Helen Hasbun, product marketing. Kiro says bots are generally available in Kiro Crew today.
Source
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
