OpenCode Message Queue Plugin

Tiny plugin that controls how user messages behave while a session is running.

Why this is useful

• Prevents mid-run interruptions: long tool runs or multi-step reasoning won't get derailed by accidental follow-ups.
• Lets you batch thoughts: drop several messages while the model is working, then deliver them in order once it finishes.
• Matches Codex-style flow: no surprise context shifts while a response is still in progress.

Modes

• immediate (default): send messages right away.
• hold: queue messages until the current run finishes, then send in order.

Visual behavior in hold

• When you send a message mid-run, the UI shows a Queued (will send after current run; N pending) placeholder.
• When the run finishes, your original message is sent once.
• A toast shows pending messages and stays until you close it (TUI only).

Install (npm/bun)

bun add opencode-message-queue
# or: npm i opencode-message-queue

Add to opencode.json:

{
  "plugin": ["opencode-message-queue"]
}

Slash command (optional)

mkdir -p .opencode/command
cp node_modules/opencode-message-queue/command/queue.md .opencode/command/

Local (no install)

This repo already includes the plugin at .opencode/plugin/message-queue.ts and the command at .opencode/command/queue.md.

Usage

Environment variables:

OPENCODE_MESSAGE_QUEUE_MODE=hold opencode
OPENCODE_MESSAGE_QUEUE_MODE=immediate opencode
# Keep queue toasts open until dismissed (default)
OPENCODE_MESSAGE_QUEUE_TOAST_DURATION_MS=0 opencode

Slash command (if installed):

/queue status
/queue hold
/queue immediate

Limitations

• Best-effort queueing: OpenCode cannot fully defer message creation yet.
• Queue is in-memory and resets when OpenCode restarts.