Signal workspace
Cases, evidence, and surface health are readable from your own systems. The schema below is the one this workspace returns.
{
"id": "KP-4821",
"threatType": "impersonation-account",
"severityScore": 94,
"relevanceScore": 0.98,
"status": "escalated",
"surface": "tiktok",
"canonicalUrl": "https://www.tiktok.com/@kestrelpay.rewards/video/7481102934",
"classifierReason": "Profile reproduces the official avatar and bio, claims to act for th...",
"harmSignals": ["credential-request","money-solicitation","official-insignia","high-reach"],
"classifierVersion": "sentinel-threat-v1",
"evidence": [
{ "kind": "screenshot", "payloadHash": "sha256:41b98070...f48a21" }
]
}Search cases, inspect a classification, and read surface health from any MCP client.
{
"mcpServers": {
"sentinely": {
"command": "npx",
"args": ["-y", "@sentinely/mcp"],
"env": { "SENTINELY_API_KEY": "sk_live_..." }
}
}
}Keys are read-only by default. Escalation and status transitions stay in the workspace, where the role model applies.
cases:readCase records, classifications, and status
evidence:readCapture metadata and payload digests
health:readPer-surface collection state and freshness