What ships
Every module, and what each one is for.
Everything below ships today and works across both channels. It is grouped the way the
platform is actually built — two engines standing on one shared core, with the dependency
running one way only. That boundary is enforced by the build, not by convention.
Conversational
Channels, language, conversation state, the flow runtime, the agent, rendering
How the agent talks to your customer
↓consumes, through one declared interface
Back office
Records, cases, queues, scheduling, questionnaires, commerce, analytics, automation
What the agent actually operates on
↓both stand on
Shared core
Identity, access, notifications, media, integrations, metering, pack lifecycle
The substrate both engines stand on
Conversational engine
How the agent talks
Channels — WhatsApp
Inbound and outbound messaging over the WhatsApp Business Platform. Text, images, documents, voice notes, location, interactive buttons, list menus, carousels and native WhatsApp forms — with webhook signatures cryptographically verified on every inbound event.
Channels — web
A chat widget you embed on your own site with one snippet, plus a shareable direct link. It runs the same pack as WhatsApp, with no separate build.
Rendering
Lists, buttons and carousels are composed once as intent, then adapted to each channel's own limits automatically. The pack never reads a channel's caps.
Language
The agent replies in the language the customer writes in, read from their own first message. No language menu, no extra model call, no per-pack configuration. Mixed input is handled.
Conversation state
Every conversation is persisted as a complete event log: customer messages, agent replies, actions taken, system events. It is both the audit record and the memory the agent reasons over.
Flow runtime
Conversations are declared as YAML flows — decisions, branches, data lookups, screens, pauses and completions. A flow can pause mid-process waiting for input and resume exactly where it stopped on the next message, even days later.
Tap routing
Free text runs the AI agent. A button or list tap is already a decision, so it routes straight to the right step with no model call at all — which is what keeps your highest-volume conversations fast and cheap.
Collect engine
Structured data capture inside the conversation — asking only for what is still missing, validating each field, and switching to a native WhatsApp form where that is the better experience.
Media
Customers can send a voice note or a photo; the platform transcribes and reads them before the agent responds, so it answers what was actually said or shown.
Back-office engine
What the agent operates on
Records (XRM)
The record engine every other module stands on. A pack declares its own entities — leads, applications, orders, assets, members — and the platform supplies storage, field validation, lifecycle stages, relationships, timeline and change events. Adding a record type needs no database work.
Casework
Case types, routing rules, service-level targets and structured closure reasons. When a conversation needs a person, the case opens carrying the full history — no repeating, no cold handover.
Worklist
The queue, assignment and SLA engine behind casework: how work reaches a team, gets claimed, escalates as it ages, and closes.
Scheduling
Resources, availability rules, real bookable slots and confirmed bookings — appointments, viewings, consultations, inspections — driven entirely from the conversation.
Surveys
Questionnaires asked inside the conversation, with several question types, response capture and rating roll-ups. Satisfaction measured at the moment of service rather than by a call-back later.
Commerce
catalog · cart · orders · payments
Show a catalog, send single- and multi-product messages, build a cart, turn it into an order and take payment — all inside the chat, instead of handing the customer to a website. Payment providers are pluggable.
Journey analytics
The customer journey as data — declared stages, the transitions between them, and analytics showing where people actually drop out. It turns “the funnel feels leaky” into a named step with a measured rate.
Automation
Rules that fire on record events: when a record reaches a stage, send a notification, create a task, schedule a follow-up, or move another record. Declared, not coded — and with no separate scheduler to operate. Standing jobs and outbound campaigns live here too.
Performance
Derived business indicators — value, conversion, satisfaction, SLA attainment and spend — calculated from what the pack already declares, with the basis of every figure traceable.
Shared core
The substrate underneath
Identity & access
Real per-user sign-in for your team, with roles, permission scopes and teams. Every back-office operation and every console screen is gated by an explicit grant, so a person sees only the records, queues and actions their role allows.
Workspace isolation
One deployment hosts many isolated workspaces — its own data, its own agents, its own users. Run several brands, business units or clients side by side without standing up separate stacks.
Operator console
Run all of it from one place: packs, conversations and their full timelines, records, cases, queues, bookings, analytics, channels and users.
Notifications & reminders
Scheduled and event-driven notifications — reminders, updates, follow-ups — reach the customer on the web widget, and on WhatsApp inside the 24-hour service window Meta allows. Business-initiated sends outside that window require an approved WhatsApp template; that path is still landing.
Operator inbox
An in-app inbox that puts in front of your team the items actually waiting on them.
Integrations
Calling out to your other systems, and being called back. Connections and credentials are held by the operator and encrypted at rest — never embedded in a pack — with field mapping so an external system's vocabulary is translated into yours.
Email
Transactional email for invitations, password resets and verification, with delivery failures handled rather than dropped.
Usage & cost
Token, message and conversation metering rolled up per workspace, so spend is visible and attributable as you grow.
Observability
Structured logging tagged with workspace, pack and request; a full per-conversation timeline your team can read; and health and readiness endpoints for your monitoring.
Pack lifecycle
A pack is published into an immutable, content-addressed artifact, then installed, upgraded or withdrawn. Two workspaces can run two different versions of the same pack at once — which is what makes staged rollout and instant rollback possible.