How we build & audit
every connector.
Connector counts mean nothing without quality. Every OneHazel integration goes through the same spec-first, AI-built, audit-gated pipeline — so operators can trust what they activate.
From vendor docs to production-ready, gated at every step
No connector hits the marketplace without passing every gate. Failures kick back to the build agent automatically — nothing ships half-finished.
Spec-first probe
We ingest the vendor's OpenAPI spec, probe every endpoint against the live API, and reconcile drift before any code is written.
AI agent build
An autonomous agent generates the connector — auth flows, retries, types, error handling — with full traceability back to the spec.
Coverage gate
Every endpoint must pass an auto-generated smoke test against the real provider. Below threshold? The build is rejected and the agent retries.
Admin review
A human reviewer inspects the diff, the audit report, and any flagged endpoints before the connector is promoted to the marketplace.
Live audit
Connectors are re-probed on a schedule. Spec drift, auth failures, or schema changes trigger an automatic re-build — before customers notice.
What operators get, per connector
A live audit report
Per-endpoint pass/fail status, last probe timestamp, response latency. Visible to every operator before they activate the connector.
Auto-generated smoke tests
Every endpoint ships with a smoke test runnable from the in-app testing console. Fire a real call before you build a workflow.
Continuous health checks
Scheduled re-probes catch spec drift the moment a vendor changes their API — before your workflows break in production.
Full execution history
Every request and response captured for every workflow run. Debug anything, defend any compliance review, replay any failure.
See the audit for any connector
Browse the marketplace and inspect the audit report on any of the 305 connectors before you activate. No card required.