Get started

Introduction

What engage-ai is, how a customer message becomes a grounded reply, and how these docs are organised.

engage-ai is an AI reply engine for business messaging. You connect a messaging account, such as a WhatsApp Business number or a Telegram bot, and load what the business knows: its products and its policies. From then on, engage-ai answers customer messages from that data only. It sends a reply when it is confident, and hands the conversation to a person when it is not.

Everything is available through a REST API. These docs explain how the system works and how to integrate with it.

What happens to a message

customer sends a message on WhatsApp or Telegram
        │
        ▼
 ① Receive      signature verified, raw event stored, 200 in under 50 ms
 ② Identify     sender resolved to a contact, conversation opened or resumed
 ③ Should we?   automation off, or a person already replying → stop, reason recorded
 ④ Retrieve     hybrid search over your products and knowledge base
 ⑤ Draft        the model writes a reply, a confidence score and its citations
 ⑥ Check        guardrails in code: banned topics, price quotes, commitments
 ⑦ Decide       send · hold as a draft for approval · escalate to a person
 ⑧ Deliver      rate limited, retried, every attempt recorded

Each step is covered in How replies work. Two rules matter most when you integrate:

No supplied facts means no answer. A reply that cites nothing from your catalogue or knowledge base is treated as a hallucination and blocked.
New channels start in review mode. The AI drafts a reply and a person approves it before it is sent. You release drafts through the inbox.

Base URL

All requests go to:

https://api.getengageai.app

Requests and responses are JSON, except the catalogue upload, which takes multipart/form-data. Every request except signup, sign-in and health checks needs a bearer token:

curl https://api.getengageai.app/auth/me \
  -H "Authorization: Bearer $ENGAGE_API_KEY"

Where to go next

If you want to	Read
Get a working organization and see a reply in a few minutes	Quickstart
Understand tokens, API keys and sessions	Authentication
Link a WhatsApp number or Telegram bot	Connect a channel
Load a catalogue from a spreadsheet or Shopify	Products and catalogue import
Add policies and FAQs, and debug what the AI reads	Knowledge base
Build an agent inbox that approves AI drafts	Inbox and approvals
Look up a specific endpoint	API reference

The concepts

Organization. Your tenant. Every product, document, conversation and credential belongs to one organization. The organization always comes from your credential, never from a request parameter, so no endpoint accepts an organizationId and one tenant cannot reach another's data.

Channel connection. A messaging account linked to your organization: one WhatsApp phone number or one Telegram bot. Each connection has its own webhook URL and its own reply mode.

Products and knowledge documents. The facts the AI answers from. Products are stored as structured records and also turned into knowledge documents, so retrieval searches one uniform collection.

Contact and conversation. A contact is a customer. A conversation is their thread with you on one channel. Messages inside it are inbound (from the customer) or outbound (from the AI or a person).

Agent profile. The AI's persona, tone, confidence threshold and guardrails. Every organization gets a default profile with conservative settings when it is created.