Skip to main content

​
Govern AI output before it reaches customers

DataVibe evaluates generated outbound after the model writes it and before your sequencer, CRM, support tool, or inbox sends it.
DataVibe is not a prompt guardrail and does not sit inside your LLM call. Your workflow sends the completed output to DataVibe before dispatch.

Run the quickstart

Create a persisted gate submission and handle every outcome safely.

Explore the API

Review authentication, endpoints, errors, quotas, and retries.

Use an SDK

Integrate with TypeScript or Python.

Connect Clay

Cross-check outbound claims against structured source evidence.

​
What the gate returns

  • BLOCKED: a hard policy violation; never send.
  • QUEUED: human review is required; hold the message.
  • APPROVED or SENT: continue according to who owns dispatch.
  • FAILED or REJECTED: stop and follow the failure or rejection workflow.
Every persisted submission is bound to an immutable policy snapshot and receives an action_id for audit and status polling.
Use POST /v1/gate/outbound for production outbound. Use POST /v1/check only for a stateless preview.
Last modified on July 23, 2026