Loading image...Kiro
  • CLI
  • Web
  • Powers
  • Enterprise
  • Pricing
  • Docs
SIGN INDOWNLOADS
Loading image...Kiro
Loading image...Kiro
Product
  • About Kiro
  • CLI
  • Web
  • Powers
  • Pricing
  • Downloads
For
  • Enterprise
  • Startups
  • Students
Community
  • Overview
  • Ambassadors
  • Discord
  • Events
  • Showcase
Resources
  • Documentation
  • Blog
  • Changelog
  • FAQs
  • Report a bug
  • Suggest an idea
  • Billing support
Social
Site TermsLicenseResponsible AI PolicyLegalPrivacy PolicyCookie Preferences
  1. Docs
  2. CLI
  3. Chat
  4. Queue Steering

Queue steering


Mid-turn steering lets you type a follow-up message while the agent is actively processing. Instead of waiting for the agent to finish or cancelling and starting over, you can send guidance that the agent incorporates into its work.

Kiro offers two follow-up modes that control when your message reaches the agent:

ModeBehaviorBest for
Steer (default)Message injected mid-turn at next tool boundaryReal-time course correction
QueueMessage buffered locally, sent after turn endsBatching multiple follow-ups

Press Ctrl+S to toggle between modes at any time.

How it works

Steer mode

  1. Type your message and press Enter while the agent is working
  2. Your message is sent to the backend queue
  3. At the next tool boundary, the message is injected into the conversation
  4. The agent sees your guidance and adjusts its approach
  5. The agent acknowledges how it handled your input

Queue mode

  1. Type your message and press Enter while the agent is working
  2. Your message is added to a local buffer (visible in activity tray)
  3. The agent finishes its current turn normally
  4. The first queued message is sent as a new prompt
  5. Process continues until the queue is empty

Injection points

In steer mode, your message is injected at:

  • Tool boundaries — after a tool completes but before the next tool starts
  • Turn end — if the agent finishes before reaching a tool boundary

Multiple messages sent in steer mode are concatenated and delivered together.

Usage

Type and press Enter while the agent is working:

bash
> Refactor the auth module [Agent starts working, reading files...] > Actually, focus on the login function first [In steer mode: message queued for mid-turn injection] [In queue mode: message added to local buffer]

Toggling modes

Press Ctrl+S to switch between steer and queue modes. When you toggle, pending messages transfer to the new mode:

  • Steer → Queue: Backend queue content moves to local buffer
  • Queue → Steer: Local buffer content is sent to backend queue

Viewing queued messages

The activity tray shows your pending messages:

  • Press Ctrl+X to expand the activity tray
  • The tab shows "Steer" (steer mode) or "Queue (N)" (queue mode with N items)
  • Press Delete or Backspace to clear the message or selected item

Cancel and redirect

Press Ctrl+C or Esc while a steering message is queued to implement "stop that and do this instead":

  1. The current operation is cancelled
  2. Your queued message is automatically sent as a new prompt
bash
> Write unit tests for the user service [Agent starts writing tests...] > Actually, let's do integration tests > [press Ctrl+C or Esc] Redirecting to: Actually, let's do integration tests [New turn starts with integration test focus]

Steer vs. queue vs. cancel

ActionWhat happensUse when
Steer (type + Enter)Agent adjusts at next tool boundaryYou want immediate course correction
Queue (Ctrl+S, then type)Messages buffer, sent after turn endsYou want to batch multiple follow-ups
Cancel (Ctrl+C or Esc)Agent stops, queued steer message sent as new promptYou want to stop and pivot immediately

Configuration

The default toggle keybinding is Ctrl+S. Customize it via:

bash
kiro-cli settings set chat.keybindings.toggleInterruptBehavior ctrl+s

Set the startup default via /settings → terminal → interrupt behaviour.

Limitations

  • Steer mode: Only one message slot — multiple messages concatenate with \n\n
  • Queue mode: Messages sent sequentially, not batched
  • Timing dependent: If the agent finishes before your message is queued, it becomes a normal prompt
  • Mode transfer: When toggling modes, concatenated steer messages split back into separate queue items

Related

  • Goal — autonomous agent loop with built-in verification
  • Responding to messages — how to interact with agent responses
  • Rewind — fork a conversation at an earlier turn
  • Settings — configure keybindings and display preferences
Page updated: June 12, 2026
Goal
In-session settings