human trust

The human trust command family exposes the Trust Observatory programmatically — public trust cards without login, plus org-private reports and webhooks when authenticated.

Public commands (no login)

List and get published trust cards:

human trust models list
human trust models get gpt-4-turbo --json
human trust models get gpt-4-turbo --cite
human trust connectors list --limit 10
human trust methodology list
human trust audits list --cursor "<next_cursor>"

Resource	Commands
Models, connectors, routes, agents, capabilities	list, get <id>
Methodology	list, get <version>
Audits	list, get <id>

Flag	Description
--json	Machine-readable JSON
--limit <n>	Page size (default 20)
--cursor <cursor>	Pagination cursor from previous response
--cite	Markdown citation block (on get)

Org-private commands (delegation required)

human login
human trust reports generate \
  --period-start 2026-05-01T00:00:00Z \
  --period-end 2026-05-31T23:59:59Z \
  --format json

human trust reports get ocr_abc123 --json
human trust webhooks list
human trust webhooks create --url https://example.com/hook --secret-ref vault:whsec
human trust webhooks delete ows_abc
human trust disputes create --card-id card_1 --claim "Incorrect safety tier"
human trust settings get
human trust settings set --contribute true

Command	Auth	Notes
reports generate|get	Delegation JWT	Org context from token
webhooks *	Delegation JWT	Org-scoped subscriptions
disputes create	Delegation JWT	Escalates to workforce review
settings get|set	Delegation JWT	Org telemetry opt-in

Use --org did:org:… to override org DID from the delegation token.

Environment

Variable	Purpose
HUMAN_API_URL	API base (default http://localhost:3001 in dev)
HUMAN_DELEGATION_TOKEN	Bearer token for org-private commands

See also: human observatory for operator publish workflows.