Skip to main content
OpenCX trains on your articles so every AI answer comes from your own content. Sync is — edits and new articles show up on the next scheduled run, no manual re-train.
Sync is configured at AI Training → Data Sources in your OpenCX dashboard, separately from the ticketing integration at Settings → Integrations.

​
Set up the sync

1

Open Data Sources

Go to AI Training → Data Sources in the OpenCX dashboard.
2

Add a Zendesk source

Click Add source → Zendesk. Enter:
FieldValue
Subdomainacme for acme.zendesk.com.
EmailAdmin email the API token belongs to.
API TokenZendesk . Can be the same token used for the Ticketing API or a separate one scoped to read.
You can reuse the API token from the Ticketing API integration. It needs permission to read Help Center articles — any Zendesk admin token works.
3

Start the initial sync

Click Start sync. The first run pulls every article — see What gets synced below. Depending on catalog size this takes minutes to a few hours. Progress is visible in the .
4

Verify articles are indexed

Once sync completes, open AI Training and search for a known article title. It should appear with Zendesk as the source.

​
What gets synced

Published, , and articles are fetched. Each is preserved as a separate entry.
Articles are tagged as public or internal based on Zendesk’s own visibility:
  • Published articles with no become public — the AI can use them in customer-facing replies.
  • Drafts and segment-restricted articles become internal — the AI uses them only in agent-facing surfaces like Assist Mode.
This mirrors Zendesk’s access model — you don’t need to rebuild it in OpenCX.

​
Incremental updates

After the initial sync, OpenCX pulls only articles whose has changed. The schedule is set on the source — hourly or daily for most teams. are filtered out automatically.
Value
Sync cursorupdated_at
ModeIncremental deduped
CadenceConfigurable per source (hourly / daily / manual)
LocalesAll locales synced; each is a separate entry

​
Re-sync and disconnect

  • Force a re-sync: click Sync now on the source row in Data Sources. Only articles changed since last run are pulled; use this after bulk edits.
  • Full rebuild: disconnect and reconnect the source. Use this only when you’ve restructured your Help Center significantly.
  • Disconnect: click Disconnect on the source. Existing indexed articles remain searchable until you delete the source — at which point all synced articles are removed from the AI’s knowledge.
Deleting a synced source removes those articles from the AI’s context immediately. The AI will start falling back to other sources until you reconnect.

Ticketing API

How the AI uses synced articles in autopilot and handoff.

Assist Mode

How rep-facing suggestions draw on your Help Center.

Troubleshooting

Sync stuck, articles missing, wrong visibility.

Training your AI

All knowledge sources and training options.