Conversation Intent Classifier

[CiaranOMahoney]

Intercom app submission: Conversation Intent Classifier

Fork:
https://github.com/CiaranOMahoney/intercom

Main repo:
https://github.com/Trac-Systems/intercom

Trac address:
trac1lmecylkpa2vdpt87ql5ucwzp6rcatsrx0cjasnuqyyy4qqyezcqqv3lr2z

Summary
Conversation Intent Classifier assigns a single predefined intent label to a conversation based on explicit keyword matching in the customer’s message.

How it works

Trigger — HTTP POST to /webhook when a conversation message is received
Input(s) used — payload.message.body text
Rule logic — deterministic keyword matching against a fixed intent map (Billing, Bug, Feature, Cancellation, How-To)
Exact output — one label: Billing, Bug, Feature, Cancellation, How-To, or Other
Runs in demo mode — no credentials required; logs detected intent and returns label in HTTP response

Intercom Integration

Operates on conversation message text
Returns a single intent label for downstream tagging or routing
Does NOT infer sentiment, priority, urgency, or perform ML-based classification

Server startup

Request response