Developers

Treat Tenvra Trust as a tool. Resolve to a candidate set, fetch the profile and evidence, read rankable and confidence before using a number, and keep the purchasing decision in the calling agent.

Contract

OpenAPI 3.1: /openapi.json
HTTP API: /api/v1/
MCP Streamable HTTP: POST /mcp (protocol 2025-03-26 / 2026-07-28). Deprecated HTTP+SSE is not implemented.
Curated machine index: /llms.txt
Embeddable score: /embed/{entity_id} (noindex; canonical is the profile)

Operations

Method	Purpose
`POST /api/v1/resolve`	Candidate set. Never a silent unique company.
`GET /api/v1/profiles/{id}`	Axes, explanation, methodology version.
`GET /api/v1/profiles/{id}/evidence`	Ledger rows for that entity.
`GET /api/v1/profiles/{id}/score-history`	Why the number moved.
`GET /api/v1/profiles/{id}/confidence`	Profile vs fact confidence.
`POST /api/v1/compare`	Side-by-side; unrankable rows warned.

Sample resolve

POST https://tenvratrust.com/api/v1/resolve
{
  "name": "McDonald's",
  "country": "GB",
  "locality": "London"
}

Sample compare

POST https://tenvratrust.com/api/v1/compare
{
  "entity_ids": [
    "tt_loc_gb_well_reviewed_cafe",
    "tt_loc_gb_mcdonalds_new_branch"
  ]
}

The cafe has a large PUBLIC sample. The new McDonald’s branch has n=2 at 5.0 — research.v0 withholds a rankable assessment. Compare must warn, not rank the 5.0 above the cafe.