Inbound Routes

Destination-side proxy for protecting SaaS AI agents you don't control — Salesforce Agentforce, HubSpot Breeze, Zendesk AI, Intercom Fin, Sierra, Decagon, Microsoft Copilot Studio.

An inbound route is the configuration that lets a SaaS-hosted AI agent send its API calls through Interven instead of directly to your real backend. Interven scans, then forwards — with stored credentials the agent never sees.

This is the only viable Interven integration for agents that run inside vendor clouds and don't accept SDK or middleware injection.

If you're integrating an agent you can modify (LangChain, your own service), use one of the framework integrations or the Gateway CLI instead. Inbound Routes are specifically for closed-runtime SaaS agents.

How it works

SaaS Agent (Salesforce, HubSpot, Zendesk, ...)
   │
   │ configured to call Interven URL
   ▼
┌──────────────────────────────────────────────┐
│  Interven Inbound Gateway                    │
│  https://api.intervensecurity.com/inbound/   │
│       <your-prefix>/<path>                   │
│                                              │
│  1. Match path to route                      │
│  2. Identify calling agent                   │
│  3. Run full policy + risk pipeline          │
│  4. Decide ALLOW / DENY / SANITIZE / REQUIRE │
│  5. Forward to real upstream with stored     │
│     credentials                              │
└──────────────────────────────────────────────┘
   │
   │ with real Bearer token
   ▼
Real upstream (Slack / GitHub / DB / API)

The agent never holds the real credential — Interven does. You revoke or rotate in one place.

The list view

/inbound-routes

Column	Notes
Path prefix	The unique segment after `/inbound/` (e.g. `sf-slack`)
Upstream URL	Where Interven forwards on ALLOW
Tool name	Optional — pinned for policy attribution (e.g. `slack`)
Agent	Default agent attribution for calls on this route
Active	Toggle to enable/disable without deleting
Calls (24h)	Volume + decision split

Creating a route

/inbound-routes/new

Fields:

Field	Required	Notes
Path prefix	yes	Letters/digits/dashes; unique per tenant; max 32 chars
Upstream URL	yes	Absolute base URL (e.g. `https://slack.com/api`); paths after the prefix are appended
Upstream headers	optional	Auth headers Interven adds when forwarding (Authorization, X-API-Key, etc.) — encrypted at rest
Tool name	optional	One of: built-in normalizer name, `custom_proxy`, or your own custom tool ID
Default agent	optional	Which agent identity gets credited for calls on this route
Allowed methods	optional	Restrict to GET / POST / etc.
Allowed path patterns	optional	Restrict paths after the prefix (e.g. `/api/v1/{customers,orders}/*`)

Save → the route URL becomes https://api.intervensecurity.com/inbound/<prefix>/.... Point your SaaS agent at this URL.

Stored credentials

Credentials added to a route's "Upstream headers" are:

Encrypted at rest with INTERVEN_CREDENTIAL_ENCRYPTION_KEY (AES-256-GCM)
Never returned in API or UI reads — only the field name is shown after save
Rotatable without disrupting in-flight requests
Audited — every credential change is logged with operator identity

To rotate, edit the route → enter the new value → save. The old credential is purged from active config immediately.

Per-platform setup guides

Platform	Guide
Salesforce Agentforce	salesforce
HubSpot Breeze	hubspot
Zendesk AI Agents	zendesk
Intercom Fin	intercom
GitHub Copilot Coding Agent	github-copilot
Microsoft Copilot Studio	microsoft-copilot
Sierra	sierra
Decagon	decagon

Decision behavior

Decision	SaaS agent sees
`ALLOW`	Real upstream's response, transparently
`DENY`	HTTP 403 with reason codes
`SANITIZE`	Real upstream's response (Interven forwarded redacted body)
`REQUIRE_APPROVAL`	HTTP 202 with `approval_id`; the next retry within 10 min auto-allows after analyst approves

Every response from an inbound route includes:

X-Interven-Decision — the decision string
X-Interven-Trace-Id — UUID for telemetry correlation
X-Interven-Reasons — comma-separated reason codes (when present)

Security model

The route is a customer-controlled URL prefix. It cannot redirect to an arbitrary destination; only the upstream URL set at config time is reachable.
Stored credentials never leave the gateway. They're injected into the outbound forward and stripped from any response surface.
Per-tenant isolation. Routes are scoped to your tenant; another tenant cannot send traffic to your inbound route URL.
Audit log captures the full request. Body, headers (minus Authorization), decision, forwarded status — searchable in Activity.

API access

GET /v1/inbound-routes — list
POST /v1/inbound-routes — create
PATCH /v1/inbound-routes/{id} — edit
DELETE /v1/inbound-routes/{id} — remove

On this page