Skip to content

Welcome to oHallo

oHallo is an autonomous customer care platform built for B2B companies. It handles inbound customer inquiries across email, chat, WhatsApp, and voice, using agents that understand your products, follow your business rules, and escalate to humans when needed.

Before diving in, here are the core ideas behind oHallo:

Workspace: Your operational environment. A workspace belongs to a brand, operates in a specific country and language, and contains all your conversations, contacts, and settings. Most companies start with one workspace and add more as they expand to new regions.

Channels: The communication paths your customers use to reach you. oHallo supports email, live chat (via an embeddable widget), WhatsApp, and voice. Each channel is connected to a workspace.

Knowledge Base: The information the assistant uses to answer questions. You add articles (Q&A pairs) about your products, services, and processes. The assistant references these when drafting replies to customers.

Policies: The business rules the assistant must follow. Policies define constraints like “quotes over EUR 5,000 require human approval” or “always require an RMA number before processing a return.” The assistant checks applicable policies before responding.

Agents: The workers that process conversations. oHallo includes system agents (orchestration, message drafting, reply validation, learning extraction) and lets you create custom specialist agents for domain-specific tasks like order lookup or returns processing.

MCP Hub: The layer that connects oHallo to your business systems. MCP (Model Context Protocol) servers connect oHallo to your ERP, CRM, order management, and product catalog. The assistant calls these tools to look up orders, check inventory, or create quotes.

Learning Loop: The feedback mechanism. When conversations are resolved, oHallo analyses them and proposes new knowledge base articles or policy rules. A human reviews and approves these proposals, so the platform continuously improves.

When a customer sends a message:

  1. oHallo receives the message through the appropriate channel
  2. The orchestrator analyses the message, classifies the intent, and determines which specialist agents to involve
  3. Specialist agents gather information from your business systems via MCP
  4. A message agent drafts a reply using your knowledge base and policies
  5. A validation agent checks the reply for accuracy
  6. The approved reply is sent back to the customer

If the assistant cannot handle a request confidently, it flags the conversation for human attention. A human can take over at any point, and the assistant resumes when released.