ZeroGrad
Capabilities · Classification

Fraud Detection

Healthy
In-context predictionEndpoint fraud-detectionProduction v7Challenger v8
Predictions / 24h
1.82M
AUC · matured outcomes
.947
Aug cohort
P95 latency
43ms
Loss avoided · Aug
$3.34M
matured outcomes
Production
Last week
v7
AUC
.947
P95
43ms
Cost / 1K
$0.18
Serving 100% of traffic
Challenger
Yesterday
v8
AUC
.956
P95
39ms
Cost / 1K
$0.16
Not serving traffic

Call it from your product

Sandbox
from zerograd import ZeroGrad

zg = ZeroGrad(remote=True)  # reads ZEROGRAD_API_KEY

result = zg.predict("fraud-detection", {
    "amount": 182.40,
    "country": "AR",
    "merchant": "electronics",
    "account_age_days": 31,
}))
Response
{
    "version": 7,
    "classes": ["legit", "fraud"],
    "predictions": [{
        "prediction": "fraud",
        "score": 0.942,
        "proba": {"legit": 0.058, "fraud": 0.942},
        "prediction_id": "prd_01J9X4K2M8"
    }]
}

Next step

v8 beats v7 on AUC (.947 → .956) in offline evaluation. Validate it on live traffic before promoting.

Activity

Today
  1. 09:42Shadow evaluation completed — v8 beat v7 on 5/5 metrics
Yesterday
  1. 18:04v8 created — new foundation model release beat v7 on held-out examples (+.009 AUC)
Monday
  1. 11:21Feature drift detected on merchant_category — severity low
Last week
  1. 16:17v7 promoted to production — blue-green completed

Use it from an agent

Every capability is also a tool. Give your agents calibrated predictions instead of letting them guess.

Private preview

One remote MCP server for the whole workspace. Works with Claude, Cursor, VS Code and any MCP client.

Claude Code
claude mcp add --transport http zerograd https://mcp.zerograd.dev/v1 \
  --header "Authorization: Bearer $ZEROGRAD_API_KEY"
mcp.json · Claude Desktop, Cursor, VS Code
{
  "mcpServers": {
    "zerograd": {
      "type": "http",
      "url": "https://mcp.zerograd.dev/v1",
      "headers": { "Authorization": "Bearer ${ZEROGRAD_API_KEY}" }
    }
  }
}
Tools exposed
  • fraud_detection_predictClassification
    Score how likely a payment transaction is fraudulent.
  • get_capability
    Health, production version and live quality metrics.
  • list_capabilities
    Every capability in the workspace.
  • record_outcome
    Send the real outcome for a prediction_id once it is known.

Tools are scoped by API key. Calls are logged, metered and show up in this capability's observability like any other request.