An AI rep that closes deals 24 hours a day

Connect your website or Instagram — AI reads your content in 30 seconds and starts replying on WhatsApp, Telegram, Instagram. Answers in 8 seconds. Never sleeps.

No signup · 30 seconds · Try it right here

Reply in8 sec
Accuracy98.8%
GDPR · Auditenterprise
0

Conversations right now

0

AI replies in 24h

0

Live channels

From signup to first message in 30 minutes

Same flow whether you have one tenant or one thousand.

Step 01

Create a workspace

Sign up at hub.arzan.cloud, pick a workspace slug, get an API key.

hub.arzan.cloud / signup

landing.howItWorks.mock1.heading

landing.howItWorks.mock1.subtitle

landing.howItWorks.mock1.workspaceName

Aliya Flowers

landing.howItWorks.mock1.email

aliya@flowers.kz

landing.howItWorks.mock1.password

••••••••••

landing.howItWorks.mock1.hint

Step 02

Connect a channel

Telegram needs only a bot token. WhatsApp / Instagram go through our reviewed Meta app — no app review on your side.

settings / channels

landing.howItWorks.mock2.connectedCount
WhatsApp Business
landing.howItWorks.mock2.statusLive
Instagram Direct
landing.howItWorks.mock2.statusLive
Telegram
landing.howItWorks.mock2.statusConnect
Messenger
landing.howItWorks.mock2.statusConnect

landing.howItWorks.mock2.hint

Step 03

Embed or call

Use the `iframe` for a hosted inbox, or the REST/Webhook API to bring your own UI.

А
landing.howItWorks.mock3.agentName
landing.howItWorks.mock3.agentMeta
landing.howItWorks.mock3.statusPublished

landing.howItWorks.mock3.systemPromptLabel

landing.howItWorks.mock3.systemPromptText

landing.howItWorks.mock3.toolsLabel

search_catalogbook_deliverylookup_orderhandoff_to_human

landing.howItWorks.mock3.knowledgeLabel

landing.howItWorks.mock3.kbFile1
landing.howItWorks.mock3.kbFile2
Step 04

Turn on the AI agent (optional)

Set a system prompt, choose a model, define tools. Hub will reply 24/7 and hand off when needed.

landing.howItWorks.mock4.inboxLabel

landing.howItWorks.mock4.thread1Name
landing.howItWorks.mock4.thread1Last
18:42
landing.howItWorks.mock4.thread2Name
landing.howItWorks.mock4.thread2Last
17:09
landing.howItWorks.mock4.thread3Name
landing.howItWorks.mock4.thread3Last
16:34
А
landing.howItWorks.mock4.thread1Name
landing.howItWorks.mock4.contactMeta
landing.howItWorks.mock4.statusAiHandled
landing.howItWorks.mock4.msg1
landing.howItWorks.mock4.msg2
landing.howItWorks.mock4.msg3
landing.howItWorks.mock4.msg4
landing.howItWorks.mock4.aiTyping
— ✦ —

See how the agent thinks, replies, and hands off

A real conversation, real RAG hit, real handoff. Not a video. Not a mock.

LIVE · Florist · WhatsApp · live demo
agent_1801kqzcf
customer

stream38ms p50tools: 3/8connected

Full-duplex streaming

Token-by-token over WebSocket. Customer sees the agent typing as it thinks.

Intelligent turn-taking

VAD with adjustable eagerness. Agent waits when the customer pauses, jumps in when they're done.

Function calling

Mid-conversation tool calls — lookup, book, transfer — without breaking the dialogue flow.

Provider agnostic

Route to Claude, GPT, Gemini, DeepSeek per-tenant. Swap in real-time via circuit breaker.

Knowledge-grounded

RAG cites the source chunk. The agent never invents prices, schedules, or facts.

Human handoff

When confidence drops or the customer asks, control transfers to an operator with full context.

Your data, AI-aware

Drag-n-drop PDF, DOCX, CSV, Notion. Auto-chunked, vector-searchable, source-cited. The agent never invents — it retrieves.

  • PDF · DOCX · CSV · Notion ingest
  • Auto-reindex on update
  • Vision search — image → product
  • Cite sources in every answer
  • Per-tenant vector isolation
  • Hybrid keyword + semantic

Frontier intelligence, your choice

Bring your own keys or use Hub Credits. Plan-based defaults, runtime fallback through a circuit breaker, and provider-aware streaming with prompt caching out of the box.

Claude Opus 4.7 logo
Claude Opus 4.7
Anthropic

Best for tool-use and long reasoning

200k ctxtoolsstreaming
GPT-5.5 logo
GPT-5.5
OpenAI

Multimodal frontier

visionaudiorealtime
Gemini 3.1 Pro logo
Gemini 3.1 Pro
Google

2M context window, fastest 1M tier

2M ctxcheapfast
DeepSeek V4 logo
DeepSeek V4
DeepSeek

Reasoning at a tenth of the cost

reasoninglow-cost
Llama 4 70B logo
Llama 4 70B
Meta

Open-source, self-host ready

openself-host
Mistral Large 3 logo
Mistral Large 3
Mistral AI

European data residency

EUmultilingual

BYOK · per-tenant · plan-based defaults · runtime fallback

One API. Three lines of code.

TypeScript and Python SDKs, idempotent by design, with WebSocket subscriptions for inbound events and a fully OpenAPI-typed surface.

Available SDKs

npm@arzan/hubpypiarzan-hubnpm@arzan/hub-react
Read the OpenAPI
Uptime
99.95%
P50 latency
38 ms
Regions
EU · USA · KZ
send-message.ts
import { Hub } from '@arzan/hub';

const hub = new Hub({ apiKey: process.env.ARZAN_HUB_KEY! });

// Send a message through any connected channel
await hub.messages.send({
  conversationId: 'cnv_a1b2c3',
  text: 'Order #1245 has shipped 📦',
});

// Subscribe to inbound — Telegram, IG, WA, all unified
hub.on('message:new', (msg) => {
  console.log(`${msg.channel}: ${msg.text}`);
});

Hub vs hiring 4 SDRs

One Hub AI agent at 10 990–29 900 ₸/mo works 24/7. Compare to an SDR team’s salary — the slider shows real savings.

Messages per month15,000
How many SDRs you'd hire5 people

Auto-calculated: 1 SDR handles 3,000 messages / mo

An SDR in Kazakhstan — 250,000 ₸/mo net × 1.30 (taxes + workplace). The AI agent plan is picked automatically (Start 10 990 / Pro 19 900 / Max 29 900 ₸).

Savings
1,614,010 ₸/ month

≈ 19,368,120 ₸ per year · Start plan

5 SDRs

1,625,000 ₸

salary + taxes

Hub

10,990 ₸

24/7 · Start

AI replies in 8 seconds — no missed leads, no nights without coverage.

Build a plan that fits you

Move the sliders to match your team — we'll pick the right plan and show the price instantly.

Operators in the team3 people
Channelsselected: 2
AI messages per month5,000

One AI message ≈ one Hub credit (≈ 150K tokens, roughly one dialog). Credits are included in your plan — adjust the slider to see how many you need.

Recommended plan
19,900 ₸/ month

Plan Pro · 19,900 ₸

The AI plan includes everything you need to launch assistants across your selected channels.

  • 3 operators in the inbox
  • 2 channels: WhatsApp · Instagram
  • AI agent with a knowledge base (included, no separate fee)
  • 5,000 AI messages / mo
  • All integrations: amoCRM, Bitrix24, Altegio, Kaspi
Connect plan

No card · Cancel in one click

An AI rep in 30 seconds

No card, no signup. Drop a URL and chat with your own agent.

Arzan Hub — AI agent + omni-inbox for Central Asia business — Arzan Hub