Skip to content

Crosscheck sources (opt-in)

Canonical tool: occam_crosscheck

Multi-source comparison / source agreement — one host fetches a URL through 2–4 local vantages (HTTP vs browser, anonymous vs session) and compares content fingerprints (blockMerkleRoot or contentHash).

Experimental, opt-in. Not consensus proof, not multi-node quorum, not proof the content is correct. The agreement verdict is unsigned computed JSON; individual vantage extracts may carry signed Receipt v1 envelopes you can verify separately.

Opt-in tool. Absent from tools/list unless the host starts with OCCAM_CONSENSUS_MCP=1. It runs 2+ full extracts per call.

When to use

  • You suspect cloaking, personalization, or access-walling (bot vs browser, anon vs logged-in).
  • Ordinary reads → occam_transcode.

Parameters

Parameter Type Default Required Description
url string — yes HTTP or HTTPS URL to cross-check
vantages string "http,browser" no Comma list: http, browser
session_profile string? null no Adds authenticated vantages per backend
focus_query string? null no Focus prune applied identically to every vantage

Returns

Success envelope:

  • verdict — consensus | divergent | access_divergent | inconclusive
  • consensus — vantages agreed on fingerprint. Agreement ≠ correctness.
  • divergent — materially different content across vantages.
  • access_divergent — some vantages walled, others succeeded.
  • Per-vantage results with optional signed receipt each
  • No signed verdict — re-derive agreement manually from vantage receipts if needed; no shipped tool re-derives the verdict cryptographically

Failure envelope: ok: false, url, failureCode, message, timestamp.

Failure codes

invalid_arguments, plus fetch-level codes when no vantage could extract. See failure codes.

Example

Call:

{ "url": "https://news.example/story", "session_profile": "news-example" }

A divergent verdict signals different fingerprints — not which vantage was "right."