Overview

Use system_of_intelligence when your system owns more of the intake experience and Klarefi provides the runtime, fact resolution, and evidence-backed case output.

In this mode, Klarefi is not the main hosted intake surface. It is the intake intelligence layer behind your workflow.

Typical Fit

• you already own the user-facing experience
• you want direct API-driven processing
• you want evidence-backed runtime outcomes without adopting the hosted intake loop

Basic Flow

1. Submit documents or inputs to Klarefi.
2. Track case progress through events or polling.
3. Read the runtime-backed case result.
4. Sync structured outcomes into your own systems.

Submit For Processing

POST /api/v1/process
Authorization: Bearer sk_live_...
Content-Type: application/json

{
  "document_url": "https://your-storage.com/claim-form.pdf",
  "case_type_id": "motor_claim",
  "idempotency_key": "process_2026_05_01_001",
  "external_case_id": "claim_12345",
  "trace_id": "trace_claim_12345"
}

Example response:

{
  "case_id": "case_abc123",
  "doc_id": "doc_def456",
  "job_id": "job_ghi789",
  "status": "queued",
  "idempotent_replay": false
}

Read Results

Use case reads and events to track progress until the case reaches a meaningful runtime state such as review-required, ready-for-handoff, or completed.

Klarefi's retrieval and multi-hop behavior are internal engine capabilities behind those results. Integrations should treat the runtime-backed case output as the primary interface.