Anti-spam: batch consecutive messages (cooldown)

Stop the AI from answering message by message: batch the customer's bursts and reply once with full context.

Jun 26, 2026

Watch this step in the video · Reply style and anti-spam

What anti-spam solves

Customers love to type in bursts: "hi", "quick question", "it's about my order", "number 1234"... four messages in a few seconds. With no protection your AI agent would reply to each one separately, burning extra tokens and firing half-baked answers before it has the full picture.

The Anti-spam feature (also called cooldown) groups those consecutive messages so the AI waits a few seconds, gathers everything, and replies once with the complete context.

How to turn it on

  1. Open AI Agents from the side menu and enter the agent.
  2. Go to the Behavior tab and scroll down to the Anti-spam card.
  3. Tick Batch consecutive messages.
  4. Configure the two fields that appear:
    • Wait (seconds) before replying: how long the bot waits after each customer message before answering. Range 1–60. A value of 5–8 seconds works well for most cases.
    • Maximum messages to batch: how many back-to-back messages it groups at most (1–20). Past that cap, it answers with what it has.
  5. Click Save.

How it works in practice

Every time a customer message arrives, the timer resets. If the customer keeps typing, the AI keeps waiting. Once they go quiet for the configured time (or the message cap is reached), the agent takes all the messages together as a single context and produces one reply.

The result: fewer half-answers, more natural conversations, and lower token usage.

Verify it works

Open the agent's Sandbox — or message it from a real channel — and fire off 3 or 4 short messages quickly in a row. You should get a single reply that accounts for all of them, not one per message.

If it doesn't work

  • It still replies to every message: confirm Batch consecutive messages is ticked and that you saved.
  • It takes too long to reply: lower the wait seconds. 60 seconds is a lot for chat; try 5.
  • It replies before the customer is done: bump the wait a bit, or raise the max messages to batch.
  • The customer sends tons of messages and info gets lost: raise the Maximum messages to batch.

Tips

  • For WhatsApp support, where people type in bursts, a 5–8 second window is usually the best balance between speed and clean batching.
  • Anti-spam doesn't replace the token budget cap; they're separate controls. Combine them to keep cost under control.

Related articles

Was this helpful?

Related articles

Message us on WhatsApp