Triggers

A connected integration is not just a place the agent can read from on demand. It is also a source of live events. When someone sends you an email, edits a Notion page, or charges a card on Stripe, OpenHuman receives that event in near-real-time and can decide whether to do something about it.

What a trigger is

A trigger is an external event published by an integration you've connected:

Integration	Example trigger
Gmail	GMAIL_NEW_GMAIL_MESSAGE
Slack	SLACK_NEW_MESSAGE
Notion	NOTION_PAGE_UPDATED
GitHub	GITHUB_ISSUE_OPENED, GITHUB_PULL_REQUEST_OPENED
Stripe	STRIPE_CHARGE_SUCCEEDED
Calendar	GOOGLE_CALENDAR_EVENT_CREATED

Trigger pipeline

third-party API → webhook → OpenHuman backend → Rust core
                                          ↓
                                   Trigger Triage
                                          ↓
         ┌─────────────┬─────────────┬────────────┐
         ↓             ↓            ↓            ↓
       drop      acknowledge     react      escalate

Triage decisions

Decision	What happens
drop	Nothing. Silently logged and discarded.
acknowledge	A short memory note is persisted.
react	Trigger Reactor runs with 1-2 tool calls.
escalate	Full orchestrator agent takes over with planning.

Configuration

• On by default. Once an integration is connected, its triggers feed into the pipeline automatically.
• Opt-out. Set OPENHUMAN_TRIGGER_TRIAGE_DISABLED to disable agent classification.
• Audit log. Every trigger is written to trigger history so you can see what arrived and what the classifier decided.

Privacy

• The third-party token lives on the backend, never on your laptop.
• Webhook is HMAC-verified by the backend before reaching your machine.
• Trigger payloads are processed by your local core.

See also

• Third-party Integrations - The catalog of services triggers come from.
• Auto-fetch - The polling counterpart for periodic data ingest.