Export Conversations from Intercom

Pull support conversations (tickets) from Intercom's REST API for downstream analysis, churn correlation, and product feedback extraction.

Prerequisites

• Intercom account with API access token (full access scope: read conversations, read contacts, read admins)
• Conversations already flowing through Intercom Messenger or email integration

API Endpoints

List all conversations (paginated)

GET https://api.intercom.io/conversations
Authorization: Bearer {access_token}
Accept: application/json

Query parameters:

• per_page: 20 (max 150)
• starting_after: cursor for pagination

Response includes: id, created_at, updated_at, state (open/closed/snoozed), priority, statistics (time_to_assignment, time_to_first_close, time_to_last_close, median_time_to_reply), contacts (the user), tags, conversation_rating.

Search conversations with filters

POST https://api.intercom.io/conversations/search
{
  "query": {
    "operator": "AND",
    "value": [
      {"field": "created_at", "operator": ">", "value": 1711929600},
      {"field": "state", "operator": "=", "value": "closed"},
      {"field": "contact_ids", "operator": "=", "value": "{contact_id}"}
    ]
  },
  "pagination": {"per_page": 50}
}

Use this to pull conversations for a specific user or time range. Filter by state, created_at, updated_at, contact_ids, teammate_ids, tag_ids.

Get a single conversation with full thread

GET https://api.intercom.io/conversations/{id}

Returns the full conversation including all conversation_parts (messages, notes, assignments, state changes). Each part has: part_type, body (HTML), author (user or admin), created_at.

Get conversation tags

Tags are returned in the conversation object under tags.tags[]. Each tag has id, name, applied_at, applied_by.

Get CSAT rating

The conversation_rating object on each conversation includes: rating (1-5), remark (customer comment), created_at, contact (who rated).

Pagination

Intercom uses cursor-based pagination. Loop until pages.next is null:

response = GET /conversations?per_page=150
while response.pages.next:
    response = GET /conversations?per_page=150&starting_after={response.pages.next.starting_after}

Rate Limits

• 1000 API calls per minute for most plans
• Use search endpoint with date filters rather than listing all conversations
• Cache results locally; only pull new/updated conversations since last sync

Error Handling

• 401: Token expired or insufficient scopes. Re-authenticate.
• 429: Rate limited. Back off exponentially (1s, 2s, 4s).
• 404: Conversation deleted. Skip and continue.

Output Schema

For downstream analysis, normalize each conversation to:

{
  "conversation_id": "123",
  "contact_id": "456",
  "contact_email": "user@company.com",
  "company_id": "789",
  "created_at": "2025-01-15T10:30:00Z",
  "closed_at": "2025-01-15T14:20:00Z",
  "state": "closed",
  "tags": ["bug", "billing"],
  "rating": 3,
  "rating_remark": "Took a while to resolve",
  "message_count": 8,
  "time_to_first_reply_seconds": 120,
  "time_to_resolution_seconds": 13800,
  "first_message_body": "I can't export my data..."
}

Store normalized data in PostHog as support_ticket_closed events or in your CRM (Attio) as activity records on the contact.