What This Does
Before your app sends money, shares keys, or runs risky actions, this service tells you whether the other agent looks safe, risky, or needs review.
Drop-in trust checks for agent workflows. Use one API key and these endpoints to log events, read trust + behavior scores, gate actions, and export audit trails.
https://clawtrustscores.com
Before your app sends money, shares keys, or runs risky actions, this service tells you whether the other agent looks safe, risky, or needs review.
Use the signup form on the homepage. You will get a key like claw_xxx. Copy it immediately (shown once, not emailed) and put it in your backend/server config as an environment variable (for example CLAWTRUST_API_KEY).
CLAWTRUST_API_KEY=INSERT_YOUR_API_KEY_HERE
Your backend sends the API call with header x-api-key. You get trust score (risk), behavior score (reliability), explanation, and signalQuality.
Use the score or preflight endpoint to allow, review, or block. Trigger this before risky actions (payment, key share, tool execution).
# backend .env file
CLAWTRUST_API_KEY=INSERT_YOUR_API_KEY_HERE
# OpenClaw skill config
{
"env": {
"CLAWTRUST_API_KEY": "INSERT_YOUR_API_KEY_HERE"
}
}
Put the key on your agent server/machine. Do not put it in frontend browser code.
Result will appear here.
This calls GET /v1/score using your key and returns both trust and behavior sections.
You can skip this on day one. Default policy works for most teams.
Preset result will appear here.
These actions call /v1/policy/presets/* and /v1/policy. Guardrail means sensitive payment/security events must come from verified integrations.
You can also set minSignalQuality (0-100) in POST /v1/policy to force review when score quality is too weak.
For high-risk actions, you can require portable credentials by setting requiredAttestations, requireAttestationsForRiskAbove, and attestationFailureDecision in POST /v1/policy.