SaaS Agent Protection
Zendesk AI Agents
Protect Zendesk AI Agents via Integration Builder pointing at Interven.
Zendesk AI Agents
Zendesk AI Agents invoke external APIs through the no-code Integration Builder โ configure HTTP method, URL, and authorization. Point the URL at Interven for full enforcement.
Setup
1. Create the inbound route in Interven
Console โ Inbound Routes โ Add route
- Path prefix:
zd-orders - Upstream URL:
https://your-shop-api.example.com/v1 - Upstream headers:
X-API-Key: your-real-shop-api-key - Tool name:
custom_proxy
2. Configure Zendesk Integration Builder
In Zendesk Admin Center โ AI Agents โ Integration Builder โ New integration:
- Method:
POST - URL:
https://api.intervensecurity.com/inbound/zd-orders/lookup - Authorization: None (Interven adds the real key)
- Body:
{ "orderId": "{{customer.orderId}}" }
3. Custom Action
Define a Custom Action in your AI Agent flow that invokes this integration. The action receives Zendesk's session parameters and forwards them through Interven to your real backend.
Behavior
| Interven decision | Zendesk action result |
|---|---|
| ALLOW | Real backend response, AI Agent continues |
| DENY | 403 โ AI Agent shows error message to customer |
| SANITIZE | Forwards redacted body, customer's PII never reaches backend |
| REQUIRE_APPROVAL | 202 โ AI Agent treats as transient error, customer waits or escalates |
Use cases
- PII redaction before backend: Customer types email/phone in chat; Interven redacts before forwarding to your CRM.
- Block writes outside business hours: Policy on
verb=WRITEafter 6pm. - Approval for high-value lookups: Require analyst approval for orders over $10,000.