Skip to main content

Tools

A tool is an action the agent can run during the chat. They’re an account-wide library: created once under Chat → WhatsApp → Tools, then attached to whichever agents need them. The Tools button in the top bar opens the library. Tick the ones this agent should have; the counter at the bottom tracks the selection.
Tools modal listing book_appointment, agendapro_scheduling, transfer_to_human, end_chat and check_order_status with their type badges

The account's tool library — tick what this agent needs

Each row shows the tool’s name, its description, and a badge for anything with special behaviour — HANDOFF, END CHAT, or the HTTP method for an API tool. The search box and the All Types filter matter once the library grows.

Types available on WhatsApp

Transfer, end call and agent swap are voice-only and don’t appear here. The chat equivalent of a transfer is handoff.

The description is the trigger

The AI decides when to use a tool by reading its description. Don’t describe what it does; describe when it should be used, and when it shouldn’t.
The type can’t be changed once the tool is created.

Integrations

There’s currently one provider available: AgendaPro.
There’s no screen for you to connect integrations yourself. The Dialtu team makes the connection. In the builder you’ll only see a status indicator — if it says Not connected, ask your Dialtu contact before going live.
Saving the agent sends the full list of attached tools. If you open the modal and remove one by accident, it’s detached on save.

Testing

The second tab of the right panel. You type as if you were the customer.
Test tab showing the draft notice, the API and integration tools run for real warning, the Test as contact selector and a conversation with the agent

Testing the draft on screen, with the real-tools warning at the top

API, integration and calendar tools run for real. A calendar tool that writes will book a real appointment. Use test data.
What is not executed for real: ending the chat, opt-out, handoff and template sends render as terminal cards, because they need a real session.
It’s free. The test panel consumes no credits, and unlike the voice one it works before you save: it tests the draft on your screen.
There’s an hourly usage limit. Test as contact defaults to Nobody in particular; picking a real contact shows how the agent responds with their memory and past conversations. Reset clears the thread and starts over.

Summary

The third tab. The Required to create checklist — name, prompt, number and routing — plus a recap of the configuration, including the AI Memory row.

Going live

Saving publishes immediately. There’s no server-side draft and no version history.
The only control is the Active / Paused indicator in the top bar.
Pausing an agent doesn’t stop its reminders. The process that sends them doesn’t check whether the agent is active. To stop it writing entirely, turn reminders off too.
For big changes to a live agent, duplicate it, work on the copy, then swap which one is active.

What doesn’t exist

  • Version history or rollback.
  • Knowledge base or document upload.
  • Analytics inside the agent — they’re under Analytics, filtered by that agent.
  • Bulk actions on the agent list.

Back to the start

Introduction to WhatsApp agents

The full path through this section.