Documentation

Quick start: Mentoring

Deploy a research-backed AI mentor in 5 steps:

1. Sign up at vantyxos.xyz and pick your plan (Starter, Pro, or Enterprise)
2. Create agent → pick a focus category (Career, Academic, Personal Development, Wellbeing, Life Skills, Safeguarding)
3. Name your agent, any name: "Layla", "Marcus", whatever fits your org
4. Embed the script on your site
5. Test the session flow, open your site, click the widget, run a full session

<!-- Step 4: Embed your mentoring agent -->
<script
  src="https://cdn.vantyxos.xyz/widget.js"
  data-agent-id="YOUR_ID"
></script>

Quick start: Support

Deploy an AI support agent in 3 steps:

1. Sign up → create a support agent
2. Connect your knowledge base or paste your FAQs directly
3. Embed the widget, one script tag and you're live 24/7

<script
  src="https://cdn.vantyxos.xyz/widget.js"
  data-agent-id="YOUR_SUPPORT_AGENT_ID"
  data-position="bottom-right"
></script>

Creating an agent

From your dashboard, click Create Agent. Choose the agent type:

// API: Create an agent
POST /api/agents
{
  "type": "mentoring",       // "mentoring" | "support"
  "name": "Layla",           // any name you choose
  "category": "career",      // mentoring only
  "greeting": "Hey! I'm Layla, your career mentor.",
  "tone": "friendly"         // professional | friendly | empathetic
}

Knowledge base selection

Mentoring agents come with pre-loaded research domains. Support agents use your custom knowledge.

Research domains (Mentoring)

Custom knowledge (Support)

Upload documentation, FAQs, or paste URLs. We chunk, embed, and index your content automatically.

POST /api/knowledge
{
  "agentId": "agent_xxx",
  "type": "url",            // "url" | "pdf" | "text"
  "source": "https://docs.yoursite.com",
  "label": "Product Documentation"
}

Agent naming & branding

You can also customise the widget accent colour, position, and language to match your brand.

Safeguarding settings

Every mentoring agent has built-in safeguarding. Crisis keyword categories are monitored in real-time:

What happens on detection:

1. LLM response is overridden, crisis resources shown instead
2. Admin notified via webhook
3. Session flagged in dashboard with full transcript
4. Helplines displayed: Childline (0800 1111), Samaritans (116 123), 988 Lifeline

Admins can configure additional keywords from the dashboard under Settings → Safeguarding.

Session structure

Every mentoring session follows the MENTOR.org EEP framework:

Focus categories

Each mentoring agent specialises in one of six categories:

Youth safeguarding

Mentoring agents working with under-18s have additional safeguarding layers: parental consent flows, age-appropriate language filters, mandatory session logging, and automatic escalation for any disclosure of harm. Based on Anna Freud Centre guidelines for digital safeguarding.

Reporting & outcomes

Track session counts, average duration, sentiment analysis, goal completion rates, and safeguarding events. Export reports as CSV or access via API. Outcome metrics aligned with OJJDP/NMRC impact frameworks.

Response configuration

Configure your support agent's tone, response length, and fallback behaviour. Set maximum response time (default: 2s). Define custom greetings and sign-off messages. Choose between concise or detailed response modes.

Escalation rules

Define when the agent hands off to a human: low confidence scores, specific topics, user request, or after N failed attempts. Configure escalation destinations (email, Slack, webhook).

// Escalation rule example
{
  "rules": [
    { "trigger": "confidence_below", "threshold": 0.6, "action": "escalate" },
    { "trigger": "keyword", "match": ["refund", "cancel"], "action": "escalate" },
    { "trigger": "user_request", "action": "escalate" },
    { "trigger": "failed_attempts", "count": 3, "action": "escalate" }
  ],
  "destination": {
    "type": "webhook",
    "url": "https://your-app.com/support/escalation"
  }
}

Custom knowledge

Upload PDFs, paste text, or provide URLs. Documents are chunked and embedded automatically. Support agents retrieve from this corpus to answer questions, no hallucination. Update your knowledge base at any time; re-indexing takes under 60 seconds.

Webhooks (Support)

Subscribe to events: conversation.started, conversation.ended, escalation.created, message.received.

Embed script

Full embed reference with all options:

<script
  src="https://cdn.vantyxos.xyz/widget.js"
  data-agent-id="YOUR_AGENT_ID"
  data-position="bottom-right"
  data-theme="dark"
  data-lang="en"
></script>

API reference

Base URL: https://api.vantyxos.xyz/v1

Auth: Authorization: Bearer vx_live_xxxxxxxxxxxx

Webhooks

All webhook events:

// Example: mentor.session.completed
{
  "event": "mentor.session.completed",
  "timestamp": "2026-03-06T10:00:00Z",
  "data": {
    "sessionId": "sess_xxx",
    "agentId": "agent_xxx",
    "category": "career",
    "duration": 840,
    "sentiment": "positive",
    "safeguarding_flags": 0
  }
}

Plans & Billing