AML Agentic Search

AccessGate AML includes agent-oriented lookup endpoints for investigation workflows that need more than a screening decision. These endpoints return structured entity context, relationship evidence, citations, and optional markdown profiles that can be used by analyst tools, case review screens, and AI-assisted investigation workflows.

Use these endpoints for evidence gathering and investigation support. Use the screening endpoints when you need a real-time CLEAR, REVIEW, or BLOCK decision.

When To Use It

Agentic search is useful when an analyst or workflow needs to answer questions such as:

  • Is this customer or counterparty directly listed?
  • Is this person related to, acting for, or connected to a listed entity?
  • What source and relationship evidence should be attached to a case?
  • Which related entities should be reviewed next?
  • What citations should be preserved for an investigation summary?

Access Pattern

Direct service calls use:

bash
https://ag-aml.runloci.com/v1/aml/...
X-Org-ID: your_aml_org_id
X-API-Key: your_aml_api_key

When AccessGate AML is enabled through the Loci platform proxy, calls use the Loci API base and tenant auth:

bash
https://api.runloci.com/aml/v1/aml/...
x-org-id: your_loci_org_id
x-api-key: your_loci_api_key

Direct AML keys must include the agent scope for agent lookup and relationship intelligence endpoints.

Core Endpoints

Endpoint Use
POST /v1/aml/agent/lookup Retrieve one rich entity profile, relationship context, citations, and optional markdown evidence.
POST /v1/aml/agent/batch Run lookup for up to 20 names in one request.
POST /v1/aml/relationships/search Search whether a name is related to a listed entity without treating it as a direct screening decision.
GET /v1/aml/entity/{id} Fetch a stored entity profile by ID.
GET /v1/aml/entity/{id}/relationships Fetch known outbound relationships for an entity.
GET /v1/aml/relationships/graph/{id} Fetch graph-shaped relationship data for visualization.
GET /v1/aml/relationships/stats Inspect relationship graph coverage and type distribution.

Typical Investigation Flow

  1. Run POST /v1/aml/agent/lookup for the customer, beneficiary, director, or counterparty name.
  2. If there is a direct match, review the entity profile, source list, confidence, and citations.
  3. If there is no direct match, inspect relationship_match to see whether the name is linked to a sanctioned or watched entity.
  4. Use include_related and include_reverse_relationships when the review needs wider graph context.
  5. Attach the returned relationship summary, citations, and relevant entity IDs to the case record.
  6. Use relationship graph endpoints when an analyst needs to expand the network around a known entity.

Example: Agent Lookup

bash
curl -X POST https://api.runloci.com/aml/v1/aml/agent/lookup \
  -H "x-org-id: org_123" \
  -H "x-api-key: loci_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Sani Abacha",
    "include_related": true,
    "include_reverse_relationships": true,
    "format": "markdown",
    "max_results": 10
  }'

A direct match returns match: true, entity metadata, citations, and any available markdown profile. A relationship-only match returns match: false with a relationship_match object describing the listed entity and relationship evidence.

Good Operating Practice

  • Treat agentic search output as investigation evidence, not an automatic final decision.
  • Preserve citations, entity IDs, relationship confidence, and source names in case notes or evidence records.
  • Use relationship matches to prioritize analyst review, especially for family members, associates, owners, controlled entities, and agents acting on behalf of listed entities.
  • Keep source coverage and update frequency visible to analysts, because relationship intelligence is only as complete as the configured data providers.