Obtain the Bearer token and the model_name for the configured triage model from the Forethought implementation team or Admin settings.
Send a POST request to the Triage API endpoint with the Bearer token in the Authorization header and a JSON body containing model_name, text_a (the ticket subject/title), and text_b (the ticket body/description).
Parse the response to extract the predicted intent label, confidence score, and any suggested routing or tagging metadata returned by the model.
Use the predicted intent to set Zendesk ticket tags or custom fields via the Zendesk API, enabling downstream triggers or views to route accordingly.
For bulk backfill scenarios, iterate over historical tickets and call the Triage API per ticket, respecting the API rate limit communicated by your implementation engineer.
Monitor API responses for low-confidence predictions and establish a threshold below which tickets should fall through to default routing rather than automated assignment.
Known gotchas
Forethought was acquired by Zendesk in March 2026; the standalone Triage API continues to function but its long-term roadmap is tied to Zendesk product integration — verify current endpoint availability before building new integrations.
The model_name is environment-specific and tied to a trained model instance; using an incorrect or stale model_name will return a 404 or an empty prediction, not a helpful error message.
The API evaluates text_a and text_b separately as subject and body — passing both fields in a single concatenated string into text_a degrades prediction accuracy compared to using both fields as intended.
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp