For Agent Builders

Ground Your Agents
In Real Private Data

Your agents need more than public web data. They need verified, structured, provenance-tracked private information — delivered at agent speed with clear pricing and citations.

Retrieval API

Not text chunks — retrieval units

Agents need something different from human search. They need compact, structured, high-confidence retrieval with provenance and permissions attached.

Every retrieval unit contains structured facts, entities, dates, amounts, and obligations — not just raw text. Plus confidence scores, pricing metadata, and citation terms.

Structured outputs

Facts, entities, obligations — machine-readable, not blob text

Citation provenance

Every result traces back to source with cryptographic verification

Permission-aware

Access controls enforced at the query layer, not bolted on after

Usage-based pricing

Pay per retrieval. Clear, predictable. No seat licenses.
retrieval_unit.json 
{
  "chunk_id": "ru_8f3a2b1c",
  "tenant_id": "tn_finserv_042",
  "modality": "document",
  "text": "Section 4.2 requires quarterly disclosure...",
  "structured_facts": [
    {
      "entity": "quarterly_disclosure",
      "type": "obligation",
      "due_date": "2026-03-31",
      "confidence": 0.94
    }
  ],
  "provenance": {
    "source": "compliance_handbook_v3.pdf",
    "page": 42,
    "hash": "sha256:9f86d08..."
  },
  "access_policy": "read_only",
  "price_per_retrieval": 0.003,
  "citation_terms": "attribute_required"
}

Integration

Built for agent workflows

Drop-in retrieval for any agent framework. REST API, SDK, and MCP-compatible.

REST API

Simple query → structured results. Standard HTTP with JSON responses. Works with any language, any framework, any agent.

Python SDK

Native Python client with typed responses, async support, and built-in retry logic. Designed for agent runtime integration.

MCP Compatible

Model Context Protocol compatible. Connect any MCP-capable agent directly to the private data retrieval layer.

Performance

Agent-grade latency

Human search can tolerate loose latency. Agents cannot.

<400ms

p50 retrieval latency

<1.5s

p95 with reranking

99.9%

uptime SLA target

Build with private data

Get early API access and start grounding your agents in real business data.