The VOX
operating manual.

VOX is a multi-tenant workspace, so the first step is creating your organization. From the landing page, choose Get started (or Create account), enter your email and a password, and complete the form. Your account becomes the owner of a new organization.

If you already have an account, sign in at /login to land on your dashboard. You can be a member of multiple organizations — switch between them from Profile when applicable.

After signing in, the left sidebar is your primary navigation. It collapses to icons on smaller widths or when you hit the chevron handle on the right edge — your preference is remembered between sessions.

• Conversations — the omni-channel chat hub: every customer thread from WhatsApp, Instagram, Messenger, web chat, and voice in one place.
• Tickets — actionable issues escalated from conversations.
• AI Agent — chat with your AI assistant by text or voice.
• Call — accept live transfers from the AI agent.
• Knowledge / Knowledge for AI — the documents and prompt that ground the AI’s answers.
• Settings — organization, users, and priorities.

Your avatar in the bottom-left opens your profile; the bell icon above it is the notifications page.

Conversations is the unified inbox for every customer thread — regardless of which channel they came in on. One list, one search, one detail view.

Channels surfaced here include:

• WhatsApp — messages from your WhatsApp Business number.
• Instagram — DMs to your business Instagram account.
• Messenger — Facebook page messages.
• Web chat — visitors using the embedded website widget.
• Voice — transcripts of AI Agent voice calls and live human transfers, captured automatically.
• Other — any additional integrations you connect later.

The page has two panes:

• List on the left — every conversation, most recent first, with a colored channel pill so you can spot the source at a glance. Use the channel filter chips above the list (All, Voice, Chat, …) to narrow down, or the search field to find a customer by name or message text.
• Detail on the right — the full transcript for the selected conversation, with the customer’s messages and the AI agent’s (or your team’s) replies in order, each timestamped.

AI Agent chats — both text and voice from the AI Agent page — are persisted here automatically, so you can audit what the assistant told customers, pull quotes for prompt tuning, or escalate a thread into a ticket when follow-up is needed.

Tickets capture work that needs to be tracked beyond a single message. The Tickets page lists every ticket in your organization with status, priority, owner, and last update.

• Create a ticket from the page header — or let the AI agent open one when it identifies an actionable request.
• Filter by status (open / in progress / resolved), priority, or assignee.
• Open a ticket to edit fields, add comments, or reassign. Comments are timestamped and attributed.

Priority labels are configurable from Settings → Priorities.

The AI Agent page is your direct line to the assistant. Two tabs sit at the top:

• Text chat — type a message, send, and stream the assistant’s reply. Your conversation is saved and appears under Conversations.
• Voice chat — press Start demo call to open a live voice session. Speak naturally; the agent replies aloud in real time.

During a voice call you can:

• Mute your microphone (the assistant keeps speaking; your input is paused).
• End the call with the red hang-up button.
• Watch the Live Transcript on the left fill in as you and the agent speak.

Answers are grounded in your Knowledge for AI prompt and any uploaded Knowledge sources. Update either to change how the agent answers.

The Call page is where human agents pick up conversations the AI escalates.

• Inbox (left pane) lists pending transfers with the caller’s identifier, urgency badge, and the AI’s summary of the conversation so far.
• Click Accept to join the call. The transcript pre-populates with everything the AI captured before escalation, so you have full context the moment you connect.
• Once connected, you can mute or hang up just like the AI Agent voice surface.

You can also place a Test call to AI from this page when no transfers are waiting — useful for sanity checks before going live.

This is the system prompt that shapes the AI’s persona, policies, and conversational style. Every AI reply — text or voice — is grounded in the contents of this page.

• Edit the prompt in the textarea and save. Changes apply to new conversations immediately.
• If you haven’t customized it, the AI falls back to a sensible default.
• There is a character limit (defaults to 8,000 Unicode characters) — keep it focused on policies and tone, and push raw factual content into Knowledge sources instead.

The Knowledge page holds the documents the AI can cite when answering. Upload PDFs, paste text, or add reference URLs; the AI retrieves relevant passages on demand (RAG).

• Add a source with the form at the top of the page.
• Each row shows the source’s title, type, and last update.
• Remove a source to take it out of the AI’s retrieval pool — old answers may still reference cached text until new conversations start.

Settings is split into a few panels.

• Organization — change the org name and review high-level details.
• Users — invite teammates by email, manage their roles, and remove members. Members with the right role can also invite others.
• Priorities — define the priority labels used in tickets (e.g. Low / Medium / High / Critical).

The Profile page (bottom-left avatar) is where you update your display name and password, and review every organization you belong to.

The Notifications page (bell icon) lists unread events — new ticket assignments, transfer requests you missed, and team activity.