What it does

source-trust-envelope takes a URL and answers the question a research agent actually has before it cites something: should I trust this. 

{
  "url": "https://en.wikipedia.org/wiki/Artificial_intelligence"
}

returns a single verdict with the reasoning behind it: 

{
  "trust_verdict": "trusted",
  "trust_score": 0.87,
  "content_authenticity": { "authenticity_score": 0.9, "text_verdict": "likely_human" },
  "corroboration": { "level": "strong", "distinct_domains": 3 },
  "domain_trust": { "available": true, "flags": [], "lookalike_risk": "none" }
}

Why

Two portfolio endpoints already touch pieces of this question, but neither answers it alone. content-authenticity-report checks whether the page's text and lead images read as AI-generated. website-trust-report checks whether the hosting domain is reputable. Neither one checks whether the actual claim is corroborated anywhere else. source-trust-envelope runs all three signals in one call and turns them into one deterministic verdict, instead of an agent making three separate calls and reasoning about the combination itself.

Degradation

Content authenticity is the required leg: if it fails, the call fails, since there's nothing trustworthy to blend without it. The corroboration search (independent-domain count for the URL's topic) and the domain-trust check are both optional legs; a rate limit or timeout on either degrades that part of the read (corroboration reports "unknown", domain trust reports "unavailable") instead of failing the whole call. The verdict thresholds are documented in the endpoint's source: "untrusted" fires on a low authenticity score or a high-risk domain, "trusted" requires human-read content, strong corroboration, and a clean domain together, and everything else lands on "mixed".

Price: $0.07.