AI-Drafted Reply Workflows with Webhooks (n8n, Make, Zapier)

Purpose

This article is for teams that want to automate Super Inbox replies using an external AI (for example Claude or ChatGPT) and a workflow tool such as n8n, Make, or Zapier.

The pattern is a round trip:

1. SuperSend sends a reply webhook when a contact replies.
2. Your automation drafts a response with AI.
3. Your automation sends the draft back to SuperSend so it goes out in the same email thread.

You can add a human approval step between draft and send.

Prerequisites

• A SuperSend account with email outreach running and replies landing in Super Inbox.
• A Webhook integration with the Reply event enabled.
• An automation tool that can receive HTTP webhooks and make HTTP requests (n8n, Make, Zapier, or custom code).
• Your SuperSend API key (required to send messages back—webhooks alone are outbound only).
• Permission to edit Campaign or Org integrations and to view Admin → System for the API key.

For webhook basics, event types, and delivery logs, see Webhooks: Testing, Delivery Logs, and Payloads and Webhooks: Replies, Bounces, and CRM Automation.

How the round trip works

sequenceDiagram
  participant Contact
  participant SuperSend
  participant Automation as n8n / Make / Zapier
  participant AI as AI (Claude, etc.)

  Contact->>SuperSend: Replies to campaign email
  SuperSend->>Automation: POST webhook (type: reply)
  Automation->>AI: Draft reply from reply.text + context
  AI-->>Automation: Suggested message
  Note over Automation: Optional human approval
  Automation->>SuperSend: POST send message API
  SuperSend->>Contact: Threaded reply from your mailbox

SuperSend handles email threading when you send through the conversation API—it uses the existing conversation history (In-Reply-To, References, provider thread IDs). You do not need to pass thread IDs from your automation.

Step 1: Configure a Reply webhook in SuperSend

1. Go to Campaign Settings → Integrations (for one campaign) or Admin → Integrations → Integration Library → Webhook (for team-wide setup).
2. Add or edit a Webhook account and paste your automation URL (for example an n8n Webhook node URL).
3. Enable only the events your workflow needs. For AI reply drafting, turn on Reply.
4. For less noise, turn Bounce and Autoresponse off if you do not want those POSTs—or filter on type in your automation (see Step 2).
5. Save the integration.
6. Use Test → next to Reply with Sample data or Real contact data to confirm your URL receives POSTs. Test payloads include "test": true—branch on this so tests do not send real emails.

Step 2: Filter for real replies in your automation

Every webhook body includes a top-level type field. Only continue your AI flow when:

type === "reply"

You can also branch on AI classification fields when present:

• reply.category — intent (for example interested, question, not_interested)
• reply.mood — sentiment (positive, negative, neutral)
• reply.confidence — classifier confidence from 0 to 1

Ignore requests where test is true.

Step 3: Get your SuperSend API key

Sending a message back requires the V2 API (not the webhook URL alone).

1. Go to Admin → System.
2. In IDs, find API Key → Copy.

Store this in your automation tool's credential store—do not hardcode it in shared workflow exports.

Step 4: Draft a reply with AI

Use the webhook payload as context for your AI step. Useful fields:

Example prompt pattern for your AI node:

You are drafting a short, professional reply to a sales email. Contact: {{first_name}} at {{company_name}}. They wrote: "{{reply.text}}". Write a concise reply (2–4 sentences). Do not invent pricing or meetings we did not offer.

Add your own brand voice, compliance rules, and forbidden phrases.

Optional — human in the loop: Insert an approval step (Slack message, email, or n8n Wait node) before sending. Many teams only auto-send when reply.category is interested or question and route everything else to a human.

Step 5: Send the draft back to SuperSend

Call the Send Message endpoint on the conversation from the webhook.

Replace {conversationId} with the conversationId from the webhook body.

Body (JSON):

{
  "message": "Your AI-drafted reply text here.",
  "sender_id": "SENDER_UUID_FROM_WEBHOOK",
  "subject": "Re: Original subject"
}

The API queues delivery asynchronously. A successful response includes status": "pending" and a job_id.

Full request and response reference: Conversation API on docs.supersend.io.

Example: n8n workflow outline

1. Webhook (trigger) — paste URL into SuperSend Webhook integration.
2. IF — {{ $json.type }} equals reply and {{ $json.test }} is not true.
3. HTTP Request (or OpenAI / Anthropic node) — draft reply using reply.text and contact fields.
4. IF (optional) — only continue if category is interested or question.
5. Slack or Gmail (optional) — approval before send.
6. HTTP Request — POST to https://api.supersend.io/v2/conversations/{{ $json.conversationId }}/messages with body from Step 5.

Make and Zapier follow the same pattern: Webhooks trigger → filter on type → AI step → Webhooks or HTTP action to SuperSend.

Alternative: In-app Auto Reply Rules

For fixed templates triggered by mood or labels (no external AI), use Super Inbox → Settings → Auto Replies. That path stays entirely inside SuperSend and does not require webhooks or an API key.

See Auto Reply Rules in Super Inbox.

Alternative: Claude with SuperSend MCP

If you use Claude Desktop, Claude Code, or Cursor with the SuperSend MCP server, Claude can list conversations and call send_conversation_message directly instead of you building the HTTP POST in n8n. That path is better for interactive drafting than high-volume unattended automation.

• Connect SuperSend MCP to Claude (and the Cursor/Codex articles in the same category).

Expected Result

• When a contact replies, your automation receives a reply webhook.
• After your AI (and optional approval) step, SuperSend sends your message in the original thread.
• The reply appears in Super Inbox on that conversation.
• The contact receives a normal threaded email from your sender mailbox.

Use Admin → Integrations → Webhook → Details → Delivery Logs or Campaign Settings → Integrations → View logs to confirm outbound webhook deliveries. Use your automation tool's execution history to confirm the send-back HTTP request succeeded.

Troubleshooting

• Issue: Webhook fires but no message is sent back.

Fix: Confirm your automation calls POST /v2/conversations/{conversationId}/messages with a valid API key, not only the inbound webhook URL. Check your tool's error output for 401 (bad key) or 400 (missing sender_id on email).

• Issue: sender_id is required for email conversations.

Fix: Include contact.Sender.id from the reply webhook in the POST body. If it is missing, list senders via the API and pick the mailbox that sent the original sequence email for that contact. See Conversation API.

• Issue: Reply sends from the wrong mailbox or fails on Gmail/Outlook.

Fix: Use the sender that already appears in the conversation thread. SuperSend restricts some provider threads to the mailbox that owns the thread.

• Issue: Test webhooks trigger AI or send test emails.

Fix: Skip flows when test is true in the webhook body.

• Issue: Automation runs on bounces or out-of-office mail.

Fix: Filter on type === "reply" and optionally disable Bounce / Autoresponse on the webhook integration. See Webhooks: Replies, Bounces, and CRM Automation.

• Issue: Duplicate replies (human and automation both send).

Fix: Add approval for all sends, or disable auto-send when a teammate may reply manually in Super Inbox. Consider branching only on high-confidence reply.category values.

• Issue: Need exact JSON field names or signing headers.

Fix: Use Test → and View Payload in webhook delivery logs, and the developer docs at docs.supersend.io — Webhooks.

Related Articles

• Webhooks: Replies, Bounces, and CRM Automation
• Webhooks: Testing, Delivery Logs, and Payloads
• Zapier and Make Setup
• Connect SuperSend MCP to Claude
• Work Conversations in Super Inbox
• Replies and Sequence Behavior

Updated on: 08/07/2026