Integration guide

At a glance

DataVibe sits between AI-generated content and your customer. You do not replace your email tool or your AI, you add one safety step in the middle.

Your AI writes a message  →  DataVibe checks it  →  Your team approves (if needed)  →  Message sends

Every check returns one of three outcomes:

Choose your setup path

Pick the option that matches your team today. You can start without code and add a developer integration later.

For developers. SDK (recommended)

The SDK is a small connector your engineering team installs once. After that, your app sends each AI-generated message to DataVibe before it reaches a customer. Your OpenAI or Anthropic keys stay on your servers, DataVibe only sees the text you choose to check.

Step 1: Install

Share these commands with your developer (TypeScript or Python):

npm install @datavibe.cc/sdk

pip install datavibe

Create an API key in the dashboard under Security → API Keys. Treat it like a password, your developer stores it in environment variables, not in source code.

Step 2: Check content before you send

Most teams use check(): your app already has AI output; DataVibe scores it and tells you whether to send, wait for approval, or stop.

TypeScript example

import { DataVibeClient } from "@datavibe.cc/sdk";

const dv = new DataVibeClient({ apiKey: process.env.DATAVIBE_API_KEY! });

// Send AI output to DataVibe before it goes to a customer
const result = await dv.check({
  content: modelOutput,
  contentType: "email", // or "agent_action" for tool / workflow steps
});

if (result.verdict === "blocked") {
  // Hard stop, fix the content and try again
  throw new Error("Blocked by policy");
}
if (result.verdict === "review_required") {
  // Held for a human, share the review link with your team
  return { status: "waiting_for_approval", reviewUrl: result.reviewUrl };
}
// safe, continue sending

Python example

from datavibe import DataVibeClient

dv = DataVibeClient(api_key="dv_live_…")

result = dv.check(
    "Agent proposed a $500 credit without manager approval",
    content_type="agent_action",
)

if result.verdict == "blocked":
    raise RuntimeError("Blocked, edit and resubmit")
if result.verdict == "review_required":
    return {"waiting": result.review_url}

Optional, generate and check in one step

If you prefer DataVibe to draft and check the message, use generateAndCheck(). Configure your AI provider in the dashboard under Settings → AI Provider.

// Optional: DataVibe drafts the message and checks it in one step
const result = await dv.generateAndCheck({
  prompt: "Write a polite follow-up for a churn-risk account",
  contentType: "email",
});
// Use result.content when the verdict is safe

Using LangChain or LangGraph? See the agent guide →

// Advanced: check AI output while it is still being generated
let sessionId: string | undefined;
for await (const chunk of openaiStream) {
  const res = await dv.streamCheck({
    content: chunk,
    sessionId,
    final: false,
  });
  sessionId = res.session_id;
  if (res.verdict === "blocked") throw new Error("Blocked mid-stream");
}
await dv.streamCheck({ content: "", sessionId, final: true });

{
  "mcpServers": {
    "datavibe": {
      "command": "npx",
      "args": ["-y", "@datavibe/mcp-server"],
      "env": { "DATAVIBE_API_KEY": "dv_live_…" }
    }
  }
}

For developers, direct API (email)

Prefer not to use the SDK? Send a single web request when you would normally click “send.” The Gate scans the message immediately and tells you whether it is safe, queued for review, or blocked.

Non-technical teams can skip this section, use Quickstart in the dashboard instead.

Where DataVibe fits

Today, AI outbound often looks like this:

AI / sales tool writes email  →  Email provider sends  →  Customer inbox

With DataVibe, you add one step:

AI writes email  →  DataVibe checks & approves  →  Email provider sends  →  Customer inbox

Your existing tools stay the same. You are adding a safety layer, not replacing your stack.

API keys (for developers)

Each request needs a secret API key (starts with dv_live_). Create one in Security → API Keys. Keys belong to your workspace, only your team can see submissions for that workspace.

1. Send a draft for review

Instead of sending email directly, post the draft to DataVibe. Think of it as “submit for safety check” rather than “send now.”

curl -X POST https://api.datavibe.cc/v1/gate/outbound \
  -H "Authorization: Bearer dv_live_YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "recipient": "[email protected]",
    "subject": "Quick question regarding your infrastructure",
    "body_html": "<p>Hi there, I noticed...</p>",
    "source_model": "gpt-4o",
    "campaign_id": "outbound_q3"
  }'

What to include in the request

2. Read the result

DataVibe replies immediately. If the message passes automated checks, it lands in your team's approval queue:

{
  "action_id": "gov_8f72c91a",
  "status": "queued",
  "policy_passed": true,
  "policy_violations": [],
  "review_url": "https://app.datavibe.cc/queue?action=gov_8f72c91a",
  "message": "Submission queued for human review. Approve at the review_url."
}

If something serious is wrong, invented pricing, guarantees, or other hard-block rules, the message is rejected on the spot and never enters the queue:

{
  "action_id": "gov_3c12d04f",
  "status": "BLOCKED",
  "policy_passed": false,
  "policy_violations": [
    {
      "rule": "pricing_hallucination",
      "severity": "BLOCK",
      "detail": "Pricing or discount claim detected: '30% off'. LLMs frequently hallucinate pricing: this cannot be auto-approved."
    }
  ],
  "review_url": null,
  "message": "Submission blocked by policy engine. Fix and resubmit."
}

3. Check send status (optional)

Developers can poll to learn when a queued message was actually sent. Most business users handle this in the approval queue instead.

curl https://api.datavibe.cc/v1/gate/outbound/gov_8f72c91a \
  -H "Authorization: Bearer dv_live_YOUR_API_KEY"

{
  "action_id": "gov_8f72c91a",
  "status": "SENT",
  "sent_at": "2026-05-12T10:43:22Z",
  "provider_message_id": "re_123abc456",
  "reviewed_at": "2026-05-12T10:43:18Z"
}

Status meanings

4. List past submissions (optional)

Developers building a custom UI or CRM sync can fetch queued and historical messages:

curl "https://api.datavibe.cc/v1/gate/submissions?status=QUEUED&limit=20" \
  -H "Authorization: Bearer dv_live_YOUR_API_KEY"

Filter by status (QUEUED, SENT, etc.). Default page size is 50.

Python example (no SDK)

A minimal drop-in if your team prefers plain HTTP over the SDK:

import requests

def send_via_gate(recipient: str, subject: str, body_html: str, campaign_id: str | None = None):
    """Submit a draft to DataVibe instead of sending email directly."""
    response = requests.post(
        "https://api.datavibe.cc/v1/gate/outbound",
        json={
            "recipient": recipient,
            "subject": subject,
            "body_html": body_html,
            "source_model": "gpt-4o",
            "campaign_id": campaign_id,
        },
        headers={"Authorization": "Bearer dv_live_YOUR_API_KEY"},
        timeout=10,
    )
    response.raise_for_status()
    data = response.json()

    if data["status"] == "BLOCKED":
        raise ValueError(f"Blocked: {data['policy_violations']}")

    print(f"Queued for review: {data['review_url']}")
    return data["action_id"]

What gets checked automatically

Every submission is scanned before it can send. Some issues block the message entirely; others flag it for a human reviewer but still allow queueing.

Need HIPAA, FINRA, GDPR, or industry-specific packs? Browse the Policy Library in the dashboard or start from the setup wizard.