Trustiry Developer Platform

Trust decisions your product can explain.

Add relationship evidence, destination context, and action-aware safety checks to software and AI agents. Every decision carries reason codes, unknowns, a policy version, and an expiry.

POST-first URL handlingFail-closed sensitive actionsVersioned decisions
Decision contract
response.json
{
  "action_allowed": false,
  "relationship_status": "unverified",
  "risk_status": "no_high_risk_signal_observed",
  "confidence_class": "medium",
  "unknowns": ["official_relationship_not_confirmed"],
  "expires_at": "2026-07-19T15:00:00.000Z"
}
Policy trustiry-policy-2026-07-17Action stopped
Platform capabilities

One decision layer, several ways to use it

Use the REST API directly, connect the MCP server, or issue scoped keys for a team.

01

Destination decisions

Check the exact destination before login, payment, upload, account creation, or form submission.

02

Relationship evidence

Separate reviewed official relationships from observed risk signals and unresolved unknowns.

03

Agent-safe policy

Receive an action-specific allow or stop result with reason codes, expiry, and policy version.

04

Official alternatives

When evidence supports one, return a reviewed destination instead of leaving the user at a dead end.

Integration flow

Place the check at the moment of consequence

Do not turn a decision into a permanent label. Ask before the action, respect expiry, and show what remains unknown.

Read the agent integration guide
  1. 1

    Describe the destination and action

    Send the URL with login, payment, upload, form submission, or account creation.

  2. 2

    Evaluate the current evidence

    Trustiry keeps relationship status, risk observations, confidence, and unknowns separate.

  3. 3

    Apply the result and preserve context

    Stop when action_allowed is false, display the reasons, and refresh after expires_at.

Quickstart

Make the first request

Issue a scoped key in the Trustiry Console, send one POST request, and treat the response as a time-bounded decision.

terminal
curl https://api.trustiry.com/v1/agent-safety \
  -H "Authorization: Bearer <tr_live_your_key>" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://example.com/checkout",
    "intended_action": "payment"
  }'
Build with evidence

Give every sensitive action a better checkpoint.

Create an API keyView reference