API reference

Capture review runs from your AI workflow.

Send Kivo the output, intended use, citations, and supplied source documents. The API returns a review-run record you can poll or subscribe to through webhooks.

POST /v1/review-runs
{
  "intended_use": "customer_facing",
  "ai_output": "According to source 1...",
  "sources": [
    {
      "title": "Refund policy",
      "media_type": "text/plain",
      "content": "Refunds may be offered..."
    }
  ]
}

Endpoint surface

The API keeps capture boring and explicit. Kivo is not a prompt playground or trace database.

POST/v1/review-runs

Create a review run from an AI output and supplied sources.

GET/v1/review-runs/{reviewRunId}

Read run status, claims, sources, and review metadata.

GET/v1/review-runs/{reviewRunId}/claims

List extracted atomic claims for a run.

GET/v1/review-runs/{reviewRunId}/brief

Return the compiled review brief when available.

Authentication

Use scoped API keys or organization-authenticated requests depending on the workflow.

Structured payloads

Inputs stay explicit: output, sources, intended use, strictness, and optional workflow metadata.

Request IDs

Every response should include or propagate a request ID for debugging and audit trails.