Skip to content
kirim.dev
MCP Server

Prompting the AI

Once the Kirim MCP server is connected (see Setup with Claude), you talk to it the same way you’d talk to a coworker. The model picks the right tool from your sentence — you don’t write the JSON payload, you don’t pass the phone_number_id, you just say what you want.

This page is a cookbook of prompts that work well, grouped by what you’re trying to do. Mix the Indonesian and English phrasings — the model handles both.

Sending messages

Section titled “Sending messages”

Plain text

Section titled “Plain text”

“Kirim WhatsApp ke +628123456789, bilang pesanannya sudah dikirim ya.”

“Send a WhatsApp to +628123456789 saying the package arrives tomorrow.”

“WA Andi di +628111222333: ‘Halo, jadwal meeting kita jam 3 sore besok.’”

Send with a template

Section titled “Send with a template”

Templates are required to start conversations outside the 24-hour customer-service window. Ask the model to look at what’s available first:

“Templates apa saja yang sudah approved di akun WhatsApp saya?”

“List my approved WhatsApp templates.”

Then send with one:

“Pakai template order_shipped ke +628123456789, isi nama ‘Andi’ dan resi ‘JNE12345’.”

“Send the welcome_message template to +628111222333 in Indonesian.”

Send media

Section titled “Send media”

“Kirim gambar ini ke +628123456789: https://example.com/promo.jpg, captionnya ‘Promo akhir bulan!’”

“Send a document to +628111222333 — link: https://example.com/invoice.pdf, filename invoice-may.pdf.”

Send interactive buttons

Section titled “Send interactive buttons”

“Send +628123456789 a CTA button: title ‘Lihat pesanan’, URL https://example.com/orders/123.”

“Kirim ke +628111222333 sebuah list dengan 3 pilihan: ‘Lacak pesanan’, ‘Hubungi CS’, ‘Lihat promo’.”

Browsing conversations

Section titled “Browsing conversations”

Recent activity

Section titled “Recent activity”

“Tampilkan 5 percakapan terbaru di Kirim saya.”

“Show me my open conversations from today.”

“Conversations mana yang masih pending lebih dari 24 jam?”

Drill into a chat

Section titled “Drill into a chat”

“Buka percakapan dengan Andi, tampilkan 20 pesan terakhir.”

“Show me the full message history for conversation cnv_WEDKY18Y8F7W26E49VMB3VX8ZF.”

“Apa pesan terakhir yang saya kirim ke +628123456789?”

Summarize a thread

Section titled “Summarize a thread”

“Ringkas percakapan dengan Andi dalam 3 kalimat — apa keluhannya dan apa yang sudah saya janjikan?”

“Summarize the last 50 messages in my conversation with Budi. What’s the customer’s main concern?”

“Translate the last 5 messages from this conversation into English.”

Searching contacts

Section titled “Searching contacts”

By phone

Section titled “By phone”

“Cari kontak dengan nomor +628123456789, kasih tahu namanya kalau ada.”

“Do I have a contact saved for +628111222333?”

By name or email

Section titled “By name or email”

“Cari kontak yang namanya mengandung ‘Andi’.”

“Find all contacts whose email is on @example.com.”

“Berapa kontak yang punya nama ‘Budi’ di akun saya?”

Multi-step workflows

Section titled “Multi-step workflows”

The model can chain tools to handle requests that span more than one step. A few real examples:

Send follow-up to recent conversations

Section titled “Send follow-up to recent conversations”

“Untuk setiap percakapan yang resolved dalam 24 jam terakhir, kirim pesan ‘Terima kasih sudah menghubungi kami, mohon nilai layanan kami: https://example.com/rating’.”

The model will:

  1. list_conversations with status=resolved filter
  2. Loop through each result
  3. send_message to each contact’s phone number

Find a contact then send

Section titled “Find a contact then send”

“Cari kontak namanya ‘Andi Wijaya’, kirim dia template appointment_reminder untuk besok jam 10 pagi.”

The model will:

  1. search_contacts with search="Andi Wijaya"
  2. Confirm which Andi you mean (if multiple match)
  3. list_templates to find appointment_reminder and inspect its variables
  4. send_message with the template populated

Audit recent sends

Section titled “Audit recent sends”

“Berapa pesan yang saya kirim hari ini? Apakah ada yang failed?”

“Show me all failed messages from the last 24 hours and explain why each one failed.”

Prompting tips

Section titled “Prompting tips”

Be specific about the sender

Section titled “Be specific about the sender”

If you have multiple WhatsApp numbers connected, name the one you want:

“Pakai akun ‘Kirim Chat’ untuk kirim ke +628123456789…”

Otherwise the model picks the first connected account it finds.

Phone-number format is flexible

Section titled “Phone-number format is flexible”

The model normalizes phone numbers before calling tools. All of these work:

  • +628123456789
  • 628123456789
  • +62 812-3456-789
  • 0812-3456-789 (if context makes the country obvious)

Ask “what can you do here?”

Section titled “Ask “what can you do here?””

“Apa saja yang bisa kamu lakukan dengan Kirim?”

The model will list the 6 tools and what each one is for — useful for new users learning the surface.

Give the model context

Section titled “Give the model context”

The more context you give, the better it picks tools. Compare:

  • “Send WA” → model has to ask follow-up questions
  • “Kirim WhatsApp ke nomor saya +628123456789, isinya ‘tes’” → one tool call, done

Confirm before risky actions

Section titled “Confirm before risky actions”

For bulk sends, deletes, or anything you can’t undo, the model usually asks for confirmation. If it doesn’t, ask it to show you the plan first:

“Sebelum kirim, tunjukkan ke siapa saja dan isi pesannya.”

What the model can’t do (yet)

Section titled “What the model can’t do (yet)”

These are out of scope for the current MCP tools. Use the dashboard or REST API instead:

  • Manage labels (create, attach, detach)
  • Manage webhook subscriptions
  • Bulk import contacts
  • Send broadcast campaigns
  • Connect new WhatsApp phone numbers
  • Manage billing / subscription

If your AI assistant needs one of these, tell us which one and your use case — we ship MCP tools demand-first.

Where to next

Section titled “Where to next”