Provenance

Name: Arkolith
Brand: Arkolith

Arkolith exposes source context at the finest granularity each current output supports. That may be an inline dataset label, a filing link, or a stored source record. Coverage is not universal, and a source label is not a complete lineage proof.

Inline provenance#

Many results include a compact provenance block that names the dataset, filing period, and applicable rights label. The exact fields vary by tool:

json

{
  "fund": "BERKSHIRE HATHAWAY INC",
  "quarter": "2026-03-31",
  "holdings": [ /* ... */ ],
  "provenance": {
    "source": "SEC EDGAR 13F",
    "filing_quarter": "2026-03-31",
    "license": "public domain"
  }
}

Read the block at its stated granularity

This example identifies the dataset and reporting period. It does not promise a stable evidence ID, retained raw bytes, a content hash, or row-level lineage.

Artifact records#

provenance.get can return stored SourceRecord entries when the caller already has the internal record type and record ID. Ordinary outputs do not consistently expose that ID, so this is not yet a universal hop from any returned row to its source artifact.

bash

curl -X POST https://arkolith.com/api/mcp \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "provenance.get",
    "arguments": {
      "record_type": "filing",
      "record_id": "clr4f9k2a0001"
    }
  }
}'

When a matching record exists, the response can include the source URL, Arkolith fetch time, parser version, optional metadata, and an optional raw hash:

json

{
  "record_type": "filing",
  "record_id": "clr4f9k2a0001",
  "provenance": [
    {
      "source": "SEC EDGAR",
      "source_url": "https://www.sec.gov/Archives/edgar/data/1067983/...",
      "fetched_at": "2026-05-16T03:11:42Z",
      "parser_version": "13f-v3",
      "raw_hash": null
    }
  ]
}

A null raw_hash means no content hash is recorded. fetched_at is when Arkolith fetched the artifact, not necessarily its exact source-publication time.

Capital Change Brief v1 is the first composed output to return stable evidence IDs inside its own evidence array and derivation manifest. Those IDs link the brief to primary SEC URLs and any available legacy source metadata; they do not make evidence IDs universal across other outputs or turn them into public provenance.get keys.

Why it matters for agents#

Source labels and filing links let an agent distinguish a reported fact from its own interpretation and give a human somewhere to verify the claim. Agents should still state the output's coverage, evidence granularity, and timestamp precision instead of treating a source label as proof of every derived claim.

Planned evidence contract#

The platform roadmap adds stable, directly resolvable evidence IDs to ordinary REST, MCP, page, and export outputs. Derived claims will link to a manifest containing every material source input and the algorithm version. Raw-byte retention and hashes will be included where source rights permit. These are planned capabilities, not guarantees of the current surface.

Licensing is part of source context

Rights labels are source-specific and do not replace the underlying terms. Some FRED series carry third-party restrictions. Congressional financial disclosures are served free pending counsel on their commercial-use restriction. Check the label and source terms before redistributing data.

← Previous

Credits & pricing

Point-in-time data