# AI platform: Agentic AI Search

Zipchat is an agentic AI platform. It searches across your data, takes real actions, and runs on every channel your customers use. This is what an agentic AI platform does that a standard chatbot cannot.

## Three things that make Zipchat agentic

### It finds the right answer, not the first match

Standard retrieval tools return the first document that matches a keyword. Zipchat's Agentic AI Search reads your live catalog, knowledge base, codebase, and integrated data sources at inference time, then reasons across all of them to find the most accurate answer for this specific customer in this specific context.

### One agent, every channel your customers use

The same AI engine runs across website chat, WhatsApp, Instagram DM, Facebook Messenger, and email. No separate bots to configure per channel. No inconsistent answers depending on where the customer asks. One agent that knows everything about your store or product, everywhere.

### It does not just answer. It acts.

Zipchat can query live order data, generate discount codes, initiate return requests, recommend products, and escalate to a human mid-conversation. These are Custom Tools: API calls the AI executes on behalf of the customer without sending them to another page or making them wait.

## How it works

### Data sources

Zipchat reads from:
- Live product catalog (Shopify, WooCommerce, Wix, BigCommerce, Magento, PrestaShop)
- Website content and knowledge base pages
- Uploaded documents, PDFs, and custom policy text
- Live Git repositories (GitHub, GitLab, Bitbucket) for Zipchat Code
- Custom API connections via Custom Tools

### Agentic AI Search

At inference time, the engine queries across all connected sources simultaneously, ranks the most relevant context for the current question, and synthesizes a response. It does not rely on a pre-indexed static snapshot. It reads live data on every query.

### Channels and actions

Supported channels: website widget, WhatsApp, Instagram DM, Facebook Messenger, email, AI Search.

Supported actions: product recommendations, discount code generation, order status lookup, return initiation, cart recovery messages, proactive outreach triggers, human handoff.

## Why this matters compared to a standard chatbot

Standard chatbots answer from a static FAQ or rule tree. They break on any question outside the rules and have no access to live catalog or order data.

Basic retrieval AI tools scrape your documentation once and answer from that snapshot. The snapshot ages every time you ship without updating docs.

Zipchat reads live sources on every query. No cold-start. No stale answers. No retrieval dead ends.

## Ecommerce mode

Connects to your store, indexes the live catalog on day one, and starts selling and supporting without a manual knowledge base. Proactive triggers initiate conversations based on browsing behavior. WhatsApp cart recovery sends sequences when a shopper abandons checkout.

## SaaS mode (Zipchat Code)

Connects to your live Git repository (GitHub, GitLab, or Bitbucket) and answers product and technical questions from the actual code, not from documentation that may be outdated. 96% answer accuracy from validated production code. 87% fewer tickets reaching engineering.

## Languages and availability

95+ languages supported with automatic detection. Available 24/7 with human handoff when the AI reaches its limit or a customer requests a person.