AGENT PROFILES

Kiro Crew

AWS · Agent orchestration · Coding agents · Agent workspaces

Persistent bots for development work: with individual memory, schedules and coordinated worker sessions on your computer or server.

Typical tasks

Availability in Europe
Added to the catalog on Dedicated profile with bot memory, operating boundaries, models, prices and sources.Details
01

What it does

6 features

Kiro · Persistent bot workspace

Product and access

Kiro Crew belongs to the Kiro family and normally uses Kiro CLI through ACP. Crew adds persistent bots and orchestration; the Kiro profile covers the shared agent core and its other interfaces.

Core features

6 features from provider sources · checked on 6 November 2026. Availability and limits are listed directly below.

  • Bots with individual responsibilities

    Named bots retain goals, a continuing chat and their own workspace.

    Source 2
  • Separate bot memory

    Each bot has its own memories that you can inspect, edit and delete.

    Source 2
  • Coordinate parallel work

    Bots start and supervise worker sessions; results and questions return to the bot chat.

    Source 2
  • Schedules and events

    Scheduled jobs, webhooks and heartbeats trigger recurring work. The Gateway must be running.

    Source 1
  • Custom dashboards

    Bots can present the state of their tasks and goals in customizable dashboards.

    Source 1
  • Tool permissions and logs

    Crew combines approval modes, policies, a sandbox and an audit log. Their effect depends on the backend and configuration.

    Source 6

Usable AI models

Fixed choice across providers

The standard Kiro CLI access offers Auto or a fixed choice of models from several providers. Bots and workers can use different models. Alternative harnesses are in Preview and are not rated here as generally available open model access.

Model families named

  • Anthropic (Claude)
  • OpenAI (GPT)
  • DeepSeek
  • Alibaba (Qwen)
  • Moonshot (Kimi)
  • MiniMax
  • Z.ai (GLM)
Model choice
Automatic or your choice
Your own API key
Not stated
Router or custom endpoint
Not stated
Local models
Not stated
Depends on
Plan, Admin approval, Region

The rating covers the generally available main access. Documented · Sources accessed on

Kiro: Models · Kiro Crew: Bots · Kiro Crew: Agent backends (Preview)

02

Does it fit my task?

2 practical tasks · 4 requirements and limits

A good fit if …

  • you want to delegate recurring development tasks to separate bots
  • you want to control memory and working state yourself
  • you want to combine schedules and events with inspectable work logs

Less suitable if …

  • you want a fully managed service without your own Gateway
  • you expect Preview backends to have identical capabilities

Editorial assessment based on provider documentation. Not a product test.

Practical tasks

Editorial examples based on documented features. No product tests performed.

01

Triage CI failures regularly

Your input
A repository, read access to CI and a clear escalation rule.
Possible result
A bot gathers findings and prepares scoped fixes.
What to check
Review the cause, diff, tests and approval before merging.
02

Maintain recurring research

Your input
A topic, permitted sources and a schedule.
Possible result
A bot updates its briefing and flags pending decisions.
What to check
Check sources and freshness; correct outdated memories.
EDITORIAL EXAMPLE

Set up a bot for recurring CI failure analysis.

Your possible prompt

Read the project rules. Investigate [error], propose a small change, and run the relevant checks. Show me the diff and any remaining uncertainties.

Adjust the task, data and permissions.

Requirements and limits

  • Requirement

    Continuous operation needs a host

    A running service or container enables work while your laptop is closed. You manage hosting and resources.

    Source 8
  • Requirement

    Kiro access and credits

    The default setup requires a signed-in Kiro CLI. Crew uses the Kiro subscription and its credits.

    Source 3
  • Limitation

    Alternative backends are in Preview

    Backend selection requires Developer Mode. Capabilities differ; Pi, for example, receives no Crew MCP tools.

    Source 9
  • Limitation

    Approvals depend on the mode

    Automatic and Autopilot do not request confirmation for each permitted call. Claude Code pre-approvals can bypass Crew checks.

    Source 6
03

Can I use it?

With requirements

Europe & regional availability

Local operation and model processing are assessed separately.

Reviewed scope: Your own host and Kiro model routing
Reviewed on:

  • With requirements

    Model routing depends on the account

    Your own EU host does not determine the model region. Individual subscriptions use the US endpoint; enterprise profiles in Frankfurt have model-dependent EU routes. According to the routing table, GPT-5.6 is processed globally through the US endpoint there too.

    Region, reason and source

    Region covered: EU/EEA, United Kingdom and Switzerland

    Documented reason: Account type, profile region and selected model determine routing.

    Official source · kiro.dev · Official source · kiro.dev

Europe and regulation explained

Accounts & permissions

Based on provider documentation. Permissions depend on the account, integration and organization policies. Review the permissions actually requested before use.

Configuration covered: Crew Gateway and standard Kiro access

Sources reviewed:

Show all 3 points
  • Documented

    Account used

    Kiro CLI is signed in separately; the dashboard requires an access token.

    Evidence · Evidence
  • Documented

    Admin approval

    Policies and profiles restrict permissions. Organizations can impose additional MCP controls.

    Evidence
  • Documented

    Data access & team sharing

    Chat channels use authorized user identities. Tools operate within the approved access.

    Evidence
04

Memory and conversation continuity

5 of 6 criteria documented

Persistent memories, separate bots and visible controls; session modes govern reading and writing.

Crew memory and private bot memories

Sources accessed:

Documentation by criterion

StorageDocumented

Storage and scope

Crew stores preferences and project context as Markdown and semantic and episodic memories in SQLite on the Gateway host.

SelectionDocumented

Selection and timing

Explicit lessons are saved immediately; consolidation runs after 30 messages or three hours of inactivity.

Edit & deleteDocumented

Inspect, edit and delete

Bot memories can be inspected, edited and deleted. Incognito reads without writing back; Temporary neither reads nor writes memory.

Access boundariesDocumented

Agent and channel boundaries

Bots have private memory. Regular channel sessions share long-term knowledge but keep separate conversation context.

RetrievalDocumented

Retrieval

Preferences are injected at session start. Semantic search combines vectors and keywords; keyword search remains available without embeddings.

New conversationsPartly documented

Session reset and compaction

Sessions survive restarts. The work ledger preserves the goal, next step and artifact references across compaction. The exact deletion boundaries of a full reset remain unspecified in the reviewed sources.

05

Cost and effort

2 plans · Prices as of 6 November 2026

Prices and included usage

Pricing checked

The software is open source. The standard model access uses Kiro credits; hosting costs are additional.

PlanPriceIncluded
Kiro FreeUSD 0 / month

50 monthly credits; limited model selection.

Kiro ProUSD 20 / user / month

1,000 monthly credits; premium models.

Price source

  • Further plans are available. Prices exclude applicable taxes; additional credits cost USD 0.04 each.

Setup

Install Crew, sign in to Kiro CLI and configure a bot with its task, tools and approvals.

Operation & oversight

Based on public provider documentation. Assessments are editorial inferences from the stated properties. Actual time requirements and reliability have not been measured.

Deployment covered: Your own Crew Gateway for ongoing work

Sources reviewed:

Editorial assessment

Suitable for supervised ongoing tasks with clear permissions; hosting, costs and results need regular review.

Show all 6 points
  • Documented

    Files and working state

    Sessions and accepted background tasks survive Gateway restarts.

    Evidence
  • Documented

    Sign-ins

    Kiro sign-in and channel credentials must remain valid; dashboard links have a limited lifetime.

    Evidence
  • Documented

    Starting further runs

    Schedules, webhooks and heartbeats can trigger further tasks.

    Evidence
  • Documented

    Human intervention

    Bots ask questions in the ongoing chat; independent work can continue.

    Evidence
  • Documented

    Detecting errors

    The System view and Crew Log show task status, tools and approvals.

    Evidence
  • Documented

    Resuming after failures

    The durable queue resumes accepted work after a Gateway restart.

    Evidence
06

Evidence and open questions

12 sources · 0 open points

Sources and evidence

Official product documentation reviewed. Workflows and assessments are editorial; hands-on product tests are outside the scope of this directory.

Editorial team: Michael Meierhoff · sources reviewed on . Provider information is not independent evidence of performance.

Read profile as Markdown

How we assess the sources

Provider

Kiro Crew is developed in the official Kiro project under Apache-2.0.

Operator
Amazon Web Services (AWS)
Organization
Provider and open-source community
Website
kiro.dev
GitHub stars
4.809 · Reviewed
License
Apache-2.0
Provider sourcesSource 1 Source 2 Reviewed
Changes to this profile 1
  • Dedicated profile with bot memory, operating boundaries, models, prices and sources.

These notes describe editorial changes. Source and pricing review dates appear in their respective sections.