Kiro bots

Kiro bots are the long-running agent shape that AWS's AI IDE Kiro made generally available in Kiro Crew on 2026-11-05. Official blog: https://kiro.dev/blog/crew-bots/ .

Unlike one-off agent sessions, each bot:

• Has a name, avatar and one ongoing chat owned by a responsibility (issue intake, build health, code review, release readiness);
• Keeps separate memory, notes and workspace so learnings don't bleed;
• Can run on a schedule (e.g. CI every 30 minutes);
• Spawns parallel worker sessions, redirects drift, and only interrupts you for decisions;
• Uses least privilege — research bots read-only, deploy bots need approval per change;
• Can pick different models for the bot vs its workers (strong planner + cheap workers).

Runs in Kiro Crew on a laptop, Mac Mini or EC2.

Positioning

The pitch is moving from "herding parallel sessions" to "managing a crew of bots" — productizing recurring responsibilities and storing judgment in bot notes. Same persistent-agent lane as Perplexity Teammates and hosted-agent products, but nested deep in Kiro's IDE workflow and dashboards.

Limits

• You still define jobs and permission boundaries;
• Tied to the Kiro Crew ecosystem — migration cost is real;
• Heavy day-one press; real-team retention TBD.

Who it's for

• Teams already on or evaluating Kiro / Kiro Crew;
• Eng leads who want named agents for triage, CI, review and release checks;
• Anyone comparing "AI coworker / hosted agent" products.

Getting started

1. Read https://kiro.dev/blog/crew-bots/ ;
2. Open Kiro Crew and create a bot in the bots workspace;
3. Pick a default or custom agent base; set goals, tools and schedules;
4. Start with one responsibility; review notes/dashboards before growing the crew.