SDKs, CLI and dashboard

SDKs#

The endpoint is one HTTP call, so any client works. The official SDKs are worth the dependency when you would otherwise write the same three things yourself: multipart uploads, one exception per refusal, and retries that never re-bill a detection unless you ask.

npm install @scam-ai/sdk     # TypeScript and Node 20+
pip install scamai           # Python 3.9+

import { ScamAI } from "@scam-ai/sdk";

const scamai = new ScamAI(); // reads SCAMAI_API_KEY
const det = await scamai.detect("suspect.jpg");
if (det.verdict === "LIKELY_AI") block(det.id);
else if (det.verdict === "ALERT") sendToReview(det.id, det.score);
else accept(det.id);

Both return the same envelope these pages document. Detections are never retried automatically. Opt in per call when you want a retry: it starts another run and may be charged, with the same sandbox and recognised duplicate exemptions described in Credits and limits. Read credits_used for the actual charge. Full reference on npm and PyPI.

Command line#

The same client as a terminal command, for scripts and coding agents. One line per check, credits and history included.

export SCAMAI_API_KEY=sk_...
npx @scam-ai/cli check photo.jpg      # verdict, score, credits used
npx @scam-ai/cli balance

Dashboard#

app.scam.ai: log in to manage API keys, watch the credit balance and usage, and open any detection's record from its history.

Support#

Integration questions go to support@scam.ai; include the id from the response or the request_id from the error. For regulated environments the same detection deploys on-prem or in your VPC. Scam AI is SOC 2 Type II audited and GDPR compliant; see the trust center.