Kiro bots go GA: long-running AI agents with memory and schedules
AWS's Kiro bots are generally available in Kiro Crew: persistent AI agents with notes, schedules and approval gates, billed in credits. Scope each bot to one job.
Kiro bots are now generally available in Kiro Crew, AWS's multi-agent workspace. A bot is a long-running AI agent that owns one area of your work (CI, issue triage, release notes) instead of one task. It keeps its own memory, runs on a schedule, and starts worker agents in parallel. The Kiro team calls it a move "from task delegation to process delegation." That is the right frame. It is also where the risk lives: an agent that runs every four hours without you is an agent you have to scope like a hire.
What actually happened
From the Kiro announcement on November 5, the Kiro pricing page and VentureBeat's coverage:
- One bot, one job. Each bot has a name, one ongoing chat, a workspace and a dashboard. It can start from Kiro's default agent or a custom agent with its own prompt, model, tools and skills.
- Memory you can edit. Each bot keeps notes on projects, decisions and open items. You can view, edit or delete what it remembers. A work log records scheduled activity.
- Schedules. You set them in chat: check CI every 30 minutes, draft release notes every Friday. The example "Heartbeat" setting runs every 4 hours, up to 6 runs a day.
- Workers. A bot starts parallel worker sessions, redirects them when they drift, and reports back. You can pick a different model and reasoning effort for the bot and for its workers.
- Approval gates. Kiro says each bot gets only the tools its job needs, and a request for more tools changes nothing until you approve it. The example review bot can request changes but cannot approve or merge. The one setting still marked Preview is "Dynamic capabilities," where a bot asks for tools on behalf of its workers.
- Where it runs. A laptop, a Mac mini or an EC2 instance. VentureBeat notes that running Crew locally does not mean the models run locally.
- Pricing. No separate bot fee is listed. Usage comes out of your Kiro plan's credits: Pro is $20 a month for 1,000 credits, Teams is $20 per seat with 500 pooled credits, and extra credits cost $0.04 each. A credit is not a fixed unit of work; it depends on the task and the model.
Why Kiro bots matter for your business
Scheduled agents are the first AI coding feature that works while nobody is looking. That's the value and the exposure in one line. A bot that checks CI every 30 minutes and starts three workers each time burns credits around the clock, and the bill tracks model choice, not headcount.
So set it up the way you'd onboard a contractor. Give each bot one job and the smallest tool list that does it. Keep merge and deploy rights with a human. Put the cheap model on the workers and the strong model on the bot that reviews them. Read its notes weekly. If a bot "remembers" a wrong decision, it will repeat it on schedule.
Start with the boring, read-only jobs: CI triage, flaky test reports, release-note drafts. Watch credit burn for two weeks. Then decide what deserves write access.
Key takeaways
- Kiro bots are GA in Kiro Crew as of November 5: persistent agents with memory, schedules and parallel workers
- Tool access is per bot; extra tools need your approval, and "Dynamic capabilities" is still Preview
- No per-bot fee: usage draws on plan credits ($20 Pro for 1,000; extra credits $0.04)
- Scheduled runs spend credits whether or not anyone is watching; meter before you scale
- Start with read-only jobs and keep merge and deploy rights with a person
Want agents that run on a schedule without running up a tab? We set up scoped coding and ops agents with approval gates and cost tracking from day one. See how we build, or run the numbers on an automated workflow first.
Sources: Kiro: Introducing bots in Kiro Crew, Kiro pricing, VentureBeat.
- #kiro
- #ai-coding-agents
- #aws
- #agent-orchestration
- #developer-tools
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
E2B passes $100M ARR: agent sandboxes are now infrastructure
E2B, which gives AI agents isolated cloud computers, says it passed $100M ARR, up from $8M a year ago, with ~200,000 teams. Isolate every agent run.
Read itDropbox buys Tigris for $350M, launches DBX Agent Cloud
Dropbox will acquire Tigris Data for ~$350M and launch DBX Agent Cloud, S3-compatible storage for AI agents. No pricing yet. Keep your storage portable.
Read it