Steering queue

When a message arrives while a session run is already streaming, OpenClaw can send that message into the active runtime instead of starting another run for the same session. The public modes are runtime-neutral; Pi and the native Codex app-server harness implement the delivery details differently.

Runtime boundary

Steering does not interrupt a tool call that is already running. Pi checks for queued steering messages at model boundaries:

1. The assistant asks for tool calls.
2. Pi executes the current assistant message's tool-call batch.
3. Pi emits the turn end event.
4. Pi drains queued steering messages.
5. Pi appends those messages as user messages before the next LLM call.

This keeps tool results paired with the assistant message that requested them, then lets the next model call see the latest user input.

The native Codex app-server harness exposes

instead of Pi's internal steering queue. OpenClaw adapts the same modes there:

• text

  Copy

  steer
  batches queued messages for the configured quiet window, then sends a single
  text

  Copy

  turn/steer
  request with all collected user input in arrival order.
• text

  Copy

  queue
  keeps the legacy serialized shape by sending separate
  text

  Copy

  turn/steer
  requests.
• text

  Copy

  followup
  ,
  text

  Copy

  collect
  ,
  text

  Copy

  steer-backlog
  , and
  text

  Copy

  interrupt
  stay OpenClaw-owned queue behavior around the active Codex turn.

Codex review and manual compaction turns reject same-turn steering. When a runtime cannot accept steering, OpenClaw falls back to the followup queue where that mode allows it.

Modes

Burst example

If four users send messages while the agent is executing a tool call:

• text

  Copy

  steer
  : the active runtime receives all four messages in arrival order before its next model decision. Pi drains them at the next model boundary; Codex receives them as one batched
  text

  Copy

  turn/steer
  .
• text

  Copy

  queue
  : legacy serialized steering. Pi injects one queued message at a time; Codex receives separate
  text

  Copy

  turn/steer
  requests.
• text

  Copy

  collect
  : OpenClaw waits until the active run ends, then creates a followup turn with compatible queued messages after the debounce window.

Scope

Steering always targets the current active session run. It does not create a new session, change the active run's tool policy, or split messages by sender. In multi-user channels, inbound prompts already include sender and route context, so the next model call can see who sent each message.

Use

when you want OpenClaw to build a later followup turn that can coalesce compatible messages and preserve followup queue drop policy. Use only when you need the older one-at-a-time steering behavior.

Debounce

applies to followup delivery, including , , , and fallback when active-run steering is not available. For Pi, active itself does not use the debounce timer because Pi naturally batches messages until the next model boundary. For the native Codex harness, OpenClaw uses the same debounce value as the quiet window before sending the batched .

Related

• Command queue
• Messages
• Agent loop