Quickstart

Send the prompt before you generate anything. Vindex checks it against your policy and answers allow, review or block. That call is the whole integration.

Five lines

Create a test key in the dashboard, under API keys. It starts with vx_test_. Put it in the first line, then run all five:

export VINDEX_KEY=vx_test_...
curl https://api.getvindex.com/v1/check \
  -H "Authorization: Bearer $VINDEX_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input": {"prompt": "a lighthouse at dusk, oil painting"}}'

That's a real decision against your default policy, logged in the dashboard under test mode.

The response

{
  "id": "dec_8c1187e9ea1a4b04a5adb40f19868d19",
  "outcome": "allow",
  "tags": [],
  "checks": {
    "minors_sexual": {
      "score": 0.0085,
      "review_score": 0.01,
      "fired": []
    },
    "adult": {
      "score": 0.01,
      "review_score": 0.01,
      "fired": []
    },
    "real_person": {
      "score": 0.01,
      "review_score": 0.01,
      "fired": []
    },
    "copyrighted_ip": {
      "score": 0.02,
      "review_score": 0.02,
      "fired": []
    }
  },
  "policy": {
    "id": "pol_cb805d8999bd453f8f2b0ffd04f05b59",
    "version": 1
  },
  "engine_version": "2026.09",
  "latency_ms": 383,
  "dry_run": false,
  "mode": "test",
  "shadow_mode": false
}

Line by line:

• id: this decision. Fetch it again with GET /v1/decisions/{id}, or ask why it came out this way with /explain.
• outcome: allow, review or block. The one field your app has to act on (Decisions).
• tags: labels your policy attached, like adult. A tag never blocks anything on its own.
• checks: one entry per check your policy runs (Checks).
  • score: how likely, from 0 to 1, the prompt is what this check looks for. Compared with the policy's act line.
  • review_score: compared with the review line. Usually the same number.
  • fired: the signals that crossed their own line. Empty on a clean prompt.
• policy: the policy and version that decided. Your default, because the request didn't name one.
• engine_version: an opaque id for the screening engine version that decided this.
• latency_ms: milliseconds from the request arriving to the decision.
• dry_run: false, so this decision counts (Dry-run).
• mode: test, because a test key made it (API keys & modes).
• shadow_mode: true while your workspace is in shadow mode, when nothing is acted on yet. Workspaces made in the dashboard start that way (Shadow mode).

In your code

curl https://api.getvindex.com/v1/check \
  -H "Authorization: Bearer $VINDEX_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input": {"prompt": "a lighthouse at dusk, oil painting"}}'

// check.mjs · Node 18+ · VINDEX_KEY=vx_test_... node check.mjs
const res = await fetch("https://api.getvindex.com/v1/check", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.VINDEX_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({ input: { prompt: "a lighthouse at dusk, oil painting" } }),
});
if (!res.ok) throw new Error(`Vindex ${res.status}: ${await res.text()}`);

const decision = await res.json();
console.log(decision.outcome); // "allow" | "review" | "block"

# check.py · pip install requests · VINDEX_KEY=vx_test_... python check.py
import os

import requests

res = requests.post(
    "https://api.getvindex.com/v1/check",
    headers={"Authorization": f"Bearer {os.environ['VINDEX_KEY']}"},
    json={"input": {"prompt": "a lighthouse at dusk, oil painting"}},
    timeout=10,
)
res.raise_for_status()

decision = res.json()
print(decision["outcome"])  # "allow" | "review" | "block"

Try it here

Paste your test key and a prompt of your own.

{
  "id": "dec_8c1187e9ea1a4b04a5adb40f19868d19",
  "outcome": "allow",
  "tags": [],
  "checks": {
    "minors_sexual": {
      "score": 0.0085,
      "review_score": 0.01,
      "fired": []
    },
    "adult": {
      "score": 0.01,
      "review_score": 0.01,
      "fired": []
    },
    "real_person": {
      "score": 0.01,
      "review_score": 0.01,
      "fired": []
    },
    "copyrighted_ip": {
      "score": 0.02,
      "review_score": 0.02,
      "fired": []
    }
  },
  "policy": {
    "id": "pol_cb805d8999bd453f8f2b0ffd04f05b59",
    "version": 1
  },
  "engine_version": "2026.09",
  "latency_ms": 383,
  "dry_run": false,
  "mode": "test",
  "shadow_mode": false
}

If it fails

• 401 unauthorized: the key is missing, mistyped or revoked.
• 422 no_default_policy: the workspace has no policy yet. Create one from a preset with POST /v1/policies and {"preset": "adult-platform"} (Policies & presets).
• 503 screening_unavailable: the screening engine couldn't be reached, so nothing was decided. Retry after the Retry-After header's seconds (Rate limits & errors).