Technical Architecture Guide • Intercom REST API v2.11 • Local CSV/NDJSON/Markdown Extraction
Why Intercom CSV Export Is Missing Message Text (And How to Export Full Transcripts Locally)
If you have ever exported conversations from Intercom's reporting dashboard to prepare a SOC2 compliance archive, migrate helpdesks, or build an AI RAG knowledge base, you likely noticed an immediate blocker: the native CSV export only includes metadata columns (Conversation ID, timestamps, assignee, tag names, SLA timers) and omits the actual customer messages, agent replies, and internal notes.
1. The Structural Reason: Why the List Endpoint Doesn't Include Transcripts
In Intercom REST API v2.11, listing conversations via GET /conversations?per_page=150 returns summary objects only. The full message thread (conversation_parts.conversation_parts, up to 500 parts per ticket) is stored as a nested entity that is only returned when you query each conversation individually:
GET https://api.intercom.io/conversations?per_page=150&starting_after={cursor}
Headers:
Authorization: Bearer <READ_ONLY_TOKEN>
Intercom-Version: 2.11
# For each conversation.id in page.conversations (N+1 Detail Fetch):
GET https://api.intercom.io/conversations/{id}?display_as=html
2. The Four Hidden Pitfalls When Exporting 10,000+ Transcripts
- Rate-Limit Windows (HTTP 429): Because every page of 150 conversations requires 150 detail calls (
N+1architecture), naive scripts hit HTTP 429 within seconds unless they monitorX-RateLimit-Remaining(< 20 threshold) and pause until theX-RateLimit-Resetepoch timestamp. - Multi-Region Endpoints: EU and AU workspaces fail with
401 Unauthorizedif queried againstapi.intercom.ioinstead ofapi.eu.intercom.ioorapi.au.intercom.io. - Browser Out-Of-Memory (OOM) Crashes: Concatenating 30,000+ HTML transcripts into a single in-memory JavaScript string or single Blob exceeds V8's UTF-16 string heap limit (~500MB–1GB) and crashes the browser tab.
- Excel UTF-8 & 32,767-Character Cell Overflow: Raw HTML tags (
<p>,<br>) and un-truncated email threads exceeding 32,767 characters corrupt rows in Microsoft Excel unless sanitized and prefixed with a UTF-8 BOM (\uFEFF).
3. How ChatVault Solves All Four Pitfalls 100% Locally
ChatVault: Conversation Transcript & CSV Exporter (for Intercom) runs inside a dedicated browser workstation tab using FileSystemWritableFileStream (Direct-to-Disk streaming) and 50-conversation IndexedDB checkpoints:
- Flushes every 50 conversations directly to disk so RAM stays under 30MB whether you export 50 or 250,000 conversations.
- Converts HTML into clean plain text and outputs Summary CSV (1 row per conversation with full transcript + separate internal notes column), Detailed CSV (1 row per message part), NDJSON (.jsonl), or Obsidian / RAG Markdown (.md).
- Zero external servers: your Read-Only Access Token and customer transcripts never leave your browser.
Test a full 500-conversation multi-turn export to Excel CSV, NDJSON, or Obsidian Markdown in 3 seconds directly in your browser.
Launch 500-Chat Interactive Demo & Workstation →Disclaimer: Independent third-party utility. Not affiliated with, endorsed, or sponsored by Intercom, Inc. • Home • Privacy Policy