Zaapi

The unified inbox connects WhatsApp, Facebook, Instagram, TikTok, LINE, Shopee, Lazada, a website chat widget, Outlook and Gmail, with Shopify and HubSpot as integrations from Pro upward. That covers messaging, marketplace, commerce and CRM systems. The agent writes as well as reads. It creates orders during conversations, and marketplace review automation posts replies back to Shopee and Lazada.

Shopify native flow triggers fire on checkout abandoned and order placed. On Advanced, webhook and HTTP actions let a flow call any external endpoint and carry on with the response. Marketplace order and customer data is pulled into the conversation for context. No connector count is published, and the integrations list stops at the named set.

Sourcezaapi.com integrations, pricing feature comparison and sales use-case pagesread 2026-09-05

Flows run several steps, branch and call outside systems, and events start them as well as messages. The Flow Builder is a no code, drag and drop canvas. From Pro up it offers advanced triggers, conditions and actions for more flexible flows. On Advanced it adds webhook and HTTP actions, so a flow can branch on a condition and call an outside endpoint mid run.

Entry points include marketplace state changes, Shopify checkout and order events and inbound webhooks. Flows mix participants. Merchants design journeys that route chats to the right agents for refunds or technical support, group and tag customers, and hand off between AI chatbots and human agents. Marketplace review automation runs as its own configured sequence. Zaapi names no runtime and does not describe retry or failure handling.

Sourcezaapi.com pricing feature comparison and AI chatbot builder guideread 2026-09-05

Merchants train the AI by uploading knowledge sources, a feature on every plan. Sellers can upload documents or add links for the AI to reference. The Enterprise tier raises knowledge base limits, so the store is sized and persists instead of being gathered per conversation. Underneath, Voyage AI handles embeddings and message processing and Unstructured Technologies handles document parsing, which together form an ingestion and embedding pipeline. The knowledge base persists, with size limits by tier, fed by parsed documents and crawled links and queried through embeddings.

Sourcezaapi.com sub-processors and pricing pages, zaapi.com AI chatbot builder guideread 2026-09-05

The AI clears first contact chats and hands off to a person when it cannot resolve them. The Flow Builder routes conversations to the right agents by topic, with refunds and technical support as Zaapi's own examples, alongside auto assignment, labeling and agent availability. Basic automations include greeting and out of hours messages that limit when the AI answers at all. The AI testing environment lets an operator see behavior before it reaches customers.

The AI deployment step is an explicit act of enabling the AI to respond to real customers, so switching it on is a deliberate decision, not a default. Zaapi names no review and approve surface, pre send queue, draft review mode, confidence threshold or per message approval. Once deployed, the AI replies to customers directly. People step in by taking over after a handoff, not by signing off before a reply is sent.

Sourcezaapi.com pricing feature comparison and AI chatbot builder guideread 2026-09-05

No certification or attestation is claimed, and SOC 2, ISO 27001, GDPR and regional certifications go unmentioned. The Enterprise tier is set apart by custom reports, a dedicated account manager, increased knowledge base limits and a flow builder consultation. Its description names enterprise security as a phrase with nothing behind it.

Seats exist and conversations can be assigned to agents, but Zaapi publishes no role model, permission set, SSO or SAML, so there is no account of who inside a merchant's organization can do what. Okta Auth0 handles user authentication for Zaapi's own login, and it is not an identity integration offered to the customer. Zaapi does publish a complete, dated sub-processor register naming nineteen processors with locations, along with a privacy policy, terms of use and a public status page. It has no trust center or security page on any subdomain.

Sourcezaapi.com sub-processors, pricing and footerread 2026-09-05

Analytics are tiered. From Pro upward they cover conversation performance, agent performance, sales performance, label usage and raw data export. Broadcast performance data comes on every plan, and Advanced adds export of full message history. A merchant can measure resolution times, team output, revenue attributed to chat and where volume is concentrating.

Zaapi names no way to inspect an individual AI run, such as which uploaded knowledge source grounded a reply, which flow branch ran, which HTTP action fired or why the agent answered as it did. Zaapi's own engineering stack uses LangSmith for AI observability and tracing, SigNoz for observability and logging and Sentry for error monitoring. Tracing of each run therefore exists inside Zaapi, but Zaapi does not say whether customers can see it.

Sourcezaapi.com pricing feature comparison and sub-processors pagesread 2026-09-05

Customer details are available on every plan inside the unified inbox, and labels and saved filters carry across conversations. Marketplace order and purchase data are pulled into the thread for context, and full message history is exportable on Advanced. A lightweight CRM consolidates chat history and purchase data into one view.

An agent, human or AI, sees who it is dealing with and what they bought. Zaapi does not say what the AI keeps between separate conversations with the same shopper, at what scope or for how long. It names no way for an operator to inspect or clear what the AI holds, or to delete it for one customer.

Sourcezaapi.com pricing feature comparison, Unified inbox sectionread 2026-09-05

Cloud hosting and storage run on Amazon Web Services in Singapore, and database services on MongoDB Atlas in Singapore. AI model services are split across the United States and Singapore through Google and Microsoft, and the model providers Anthropic and OpenAI sit in the United States.

A customer can see exactly where data goes, which matters for a seller base in Southeast Asia. The arrangement is fixed. There is no region selector, VPC, single tenant, on premises or customer tenant option. The Enterprise tier adds custom reports, a dedicated account manager and increased knowledge base limits, not deployment choices.

Sourcezaapi.com sub-processors and pricing pagesread 2026-09-05

Merchants build their own agent and flows. The product ships one AI Agent that a merchant trains, tests and deploys, plus a Flow Builder in which they compose their own journeys. Zaapi's guide walks a seller through building a chatbot from scratch, setting response triggers by keyword or button and routing chats to the right agents.

Some ready made behaviors come with it, including greeting messages, out of hours responses, conversation closing messages, marketplace review automation and Shopify checkout abandoned flows, so a merchant does not start from nothing. There is no template gallery, flow library, named prebuilt agent set or entitlement for a customer to browse and adopt. Instead Zaapi offers services. Under its AI Success Kit, Zaapi builds the AI for the merchant across their channels, and Enterprise includes a flow builder consultation.

Sourcezaapi.com pricing, AI chatbot builder guide and AI services pagesread 2026-09-05

Marketplace triggers fire on review received, order created, shipped, delayed, delivered, failed, canceled and returned or refunded. Shopify native flow triggers add checkout abandoned and order placed. Webhook event triggers let any external system start a flow. Broadcasts carry smart scheduling, so work can also start on a schedule.

The events are changes in commerce state, not message arrivals, so work reaches the agent when something happens in the merchant's business. Each channel is a separate inbound stream, covering WhatsApp, Facebook, Instagram, TikTok, LINE, Shopee, Lazada, a website chat widget, and Outlook and Gmail.

Sourcezaapi.com pricing feature comparison, Automations and Unified inbox sectionsread 2026-09-05

Four model providers run underneath, and the merchant picks none of them. Google supplies cloud infrastructure and AI model services, including Gemini models. Microsoft hosts AI models through Azure AI Foundry, Anthropic and OpenAI are AI model providers, and Voyage AI handles embeddings and message processing. Zaapi lists all of them in its current and complete sub-processor register.

There is no model selector, routing rule, per agent assignment or bring your own key path. The AI Agent's settings cover knowledge sources, testing and deployment, not models. AI usage is metered as tokens priced per thousand messages, not per model, which fits with Zaapi making the choice.

Sourcezaapi.com sub-processors page, pricing pageread 2026-09-05

Webhook event triggers and webhook and HTTP actions come on the Advanced and Enterprise tiers, and Webhook events sit among the integrations beside HubSpot and Shopify. An external system can start a Zaapi flow, and a Zaapi flow can call an external endpoint, both set up in the Flow Builder without code. Zaapi publishes no API or SDK for its own platform, and no developer portal, endpoint reference, authentication model, client library or MCP server. Its help center carries operational articles, not developer documentation.

Sourcezaapi.com pricing feature comparison, Automations and Integrations sectionsread 2026-09-05

The AI Agent follows a three stage lifecycle on every plan. AI training means uploading knowledge sources. AI testing means trying the AI in a secure environment. AI deployment means enabling the AI to respond to real customers. The testing stage separates a rehearsal environment from live customers, so a merchant tries a change before it reaches anyone.

All three stages are available on Basic, Pro, Advanced and Enterprise. Zaapi advises sellers to test a flow internally and fix any issues before rolling out a chatbot. Zaapi names no scored test set, pass rate, regression suite or published threshold, and no experiment or holdout after deployment.

Sourcezaapi.com pricing feature comparison, AI Agent sectionread 2026-09-05

Every outside system is reached through credentialed channel connections and integrations. They include marketplace APIs for Shopee and Lazada, Meta and TikTok platform connections, LINE and WhatsApp business channels, Shopify and HubSpot integrations, and webhook and HTTP actions for anything else. Zaapi's own surfaces are a chat widget the merchant embeds, a web app and a mobile app.

Marketplace review automation posts replies onto Shopee and Lazada review pages, which Zaapi does not own, but it does so through the marketplace integration and does not navigate an interface. Zaapi names no browser, desktop or remote computer control, and no session or interface automation.

Sourcezaapi.com pricing feature comparison and integrations pagesread 2026-09-05