Everything your front desk does, around the clock
Answer by chat or out loud, qualify, book, and hand off — a complete lead loop from a single dashboard.
A receptionist that closes the loop
When a visitor shows interest, the bot collects name, phone, project, ZIP, and timeframe conversationally — or through a quote form. Choose from 21 industry templates (concrete, HVAC, med spa, dental, law, real estate…) modeled on how each business qualifies a first call, then tweak the questions or add your own.
- 21 industry question templates
- Custom questions per bot
- ZIP-based service-area screening
- Email alerts, lead statuses & CSV export
From question to booked appointment
List your services once — name, price, duration — or import them straight from your website or booking page. The bot quotes real prices, suggests the right service, and takes the booking request: what, when, and how to reach them. You confirm with one click; both sides get an email. Already using Vagaro, Calendly, or another scheduler? The bot hands visitors to your booking page instead.
- Prices quoted only from your service list
- Requests you confirm — the bot never overpromises a time
- Keep your existing scheduler: handoff mode links visitors to it
- Zapier webhook: push bookings to your calendar or CRM
Some visitors would rather just talk
A "Talk to us" button in the widget opens a real spoken conversation in the browser — no phone number to buy, no app, nothing for the visitor to install. It answers from the same services and knowledge as the chat, reads phone numbers and ZIP codes back to confirm them, and captures the lead or takes the booking while it talks. Every call becomes a transcript, a one-line summary, and a conversation in your inbox.
- Rides the widget you already installed
- Confirms digits out loud before saving them
- Transcript, summary and lead from every call
- Visitors told it is an AI before the mic opens
Teach it what you already wrote
Add page URLs, a whole sitemap, PDF / DOCX / CSV files, hand-written FAQs, or pasted text. Mazingi extracts the readable text, splits it into chunks, and indexes every chunk with Gemini embeddings — so answers come from meaning, not just matching words.
- Single pages or sitemap.xml crawling
- PDF, DOCX, TXT, MD, CSV uploads
- Re-crawl and retry with one click
- Per-source status and error reporting
Citations on every reply
Answers are generated only from the chunks retrieved for that exact question, and each reply carries the sources it used. When nothing relevant exists, the bot says so with your fallback message — and the miss lands in your gap report.
- Strict knowledge-base mode
- Source chips on every answer
- Fallback responses you control
- Unanswered-question tracking
Your team, one click away
Every widget conversation streams into the dashboard inbox live. An agent can pause the AI, take over the thread, leave internal notes, and resolve — the visitor stays in the same chat window the whole time.
- Real-time inbox
- AI pause / takeover per conversation
- Internal notes and statuses
- Thumbs-up / thumbs-down feedback
One tag, any platform
The widget ships as one async script tag rendered inside a Shadow DOM. Host-page CSS cannot break it; it cannot break your page. Your Gemini key never leaves the server — the tag carries only a public bot ID guarded by domain allowlists and rate limits.
- WordPress, Shopify, Webflow, React, HTML
- Shadow DOM isolation
- Per-bot domain allowlist
- Server-side API keys only
Looks like your product
Brand color, launcher position, greeting, header title, and dark mode are configurable per bot. Preview changes in the dashboard before you publish them.
- Custom colors and greeting
- Bottom-left or bottom-right launcher
- Dark mode
- Mobile-ready layout
Numbers you can act on
Conversations, resolution rate, escalations, response latency, token usage, and feedback — filtered by date range and by bot. Every metric is computed from your real data.
- Per-bot and date filters
- Resolution and escalation rates
- Token usage tracking
- Knowledge gap reports
Tenants stay tenants
Every request is verified twice — your session, then your workspace membership and role — before any data is touched. Retrieval is scoped to a single bot, so cross-tenant leakage is structurally impossible.
- Per-workspace data isolation
- Role-based access (owner / admin / agent / viewer)
- Rate limiting and input validation
- Prompt-injection hygiene
See it answer your questions
Set up a workspace and index your first source in minutes.