WhatsApp broadcasts by chatting with your own AI assistant.

You already have an AI assistant. Now it can drive WACO: "send to customers labelled member: Hi {Nama}, weekend promo 20% off until Sunday" — and the agent handles the WhatsApp template, Meta approval, the recipient list, a preview, then sends once you say go. Connected through the Model Context Protocol, the open standard used by Claude, ChatGPT, Cursor, and n8n. Sign in once with your WACO account, no key copying, no code.

Get started See pricing

What it feels like

One conversation with your AI assistant — from request to report. No templates, variables, or categories to learn.

Send to customers labelled member: Hi {Nama}, weekend promo 20% off until Sunday.

Claude · WACO MCPTemplate submitted to Meta and approved. The member label has 412 contacts. Sample: "Hi Budi, weekend promo 20% off until Sunday." Estimated cost 412 × Meta marketing rate. Send now?

OK, send it.

Claude · WACO MCPSent to 412 numbers. Report so far: 389 delivered, 241 read, 12 replied.

What your agent can do

💬

Broadcast from plain text

Write the message as you would to a customer; per-person parts in {Nama} or {Tagihan}. WACO builds the template, submits it to Meta, and the agent tracks approval. You never need to know what a template is.

🏷️

Database & labels from chat

Paste a list of 300 customers to the agent and it files them into Contacts under one label. Then "send to everyone with that label" — variables fill in from names and your columns. Image headers by URL.

🛡️

Never sends without consent

The agent must show the recipient count, three filled-in samples, and the estimated cost. Sending only works with that preview token, 30 minutes, after you approve.

📊

Per-person reports & message log

"Who failed, and why?" — a per-number list with the reason and its meaning; the status of one message (delivered, read) and a 30-day history for any number, whenever you ask.

🔐

OAuth sign-in, revoke any time

Add the connector in claude.ai or ChatGPT, sign in to WACO, allow. Each app gets its own key shown under Developer — revoke one without cutting off the others.

🧩

Included in the plan

No extra cost. The agent runs in your own app; WACO only provides the socket. The same tools are also available as a REST API.

Setting it up

Claude.ai: Settings → Connectors → Add custom connector → URL https://waco.id/mcp → Connect → sign in to WACO → Allow. ChatGPT works the same through its Connectors menu. Claude Code, Cursor, and n8n use an API key. Details in the MCP guide.

Questions about this

Can the AI assistant send a broadcast without me knowing?

No. Sending only works through a preview token created after the agent shows the summary, and the tool requires an explicit user-approval flag. It is also limited to five sends per minute per key.

Is my data used to train AI?

WACO runs no model for this feature; the agent runs in the service you choose (Claude, ChatGPT, etc.) under its own data policy. WACO only answers the tool requests the agent sends.

Which clients are supported?

Anything that speaks remote MCP (Streamable HTTP): claude.ai and Claude Desktop, ChatGPT, Claude Code, Cursor, Windsurf, n8n. OAuth-capable clients just sign in; others use an API key.

More questions? See the WACO guides or say hello through the chat bubble.