Everything a website agent needs, nothing it should make up.
Each feature has its own page with an interactive demo. Status is marked honestly: everything below is built unless it says otherwise.
- 01
Grounded answers
Answers only from your knowledge base, never from guesswork.
Every answer is built from facts retrieved from your knowledge base. When nothing relevant is on record, the agent says so, offers a person or an email follow-up, and logs the question as a gap.
Explore - 02
Structured agent actions
Tool calls, not keyword matching, for handoff, leads and CTAs.
The agent acts through typed tool calls that the backend validates before anything happens. A model that produces bad arguments is ignored, not obeyed.
Explore - 03
Human handoff & phone takeover
Any teammate can claim a live chat from their phone.
When a visitor asks for a person, raises an always-human topic, or the agent can't help, the conversation goes to the team inbox. A push reaches every online member with access to the project, and the first to tap Assign to me takes it.
Explore - 04
Point & visual feedback
Visitors point at the page and send a screenshot.
Visitors can click an element or drag a region anywhere on the page. The widget captures that area in the browser, masks sensitive fields, and attaches it to the message or files it as feedback.
Explore - 05
Embeddable widget
One script tag, five layouts, 17.5 KB gzipped.
Exploremango.jsis vanilla TypeScript in a shadow DOM, so it works on any site stack without style clashes. Every project starts from the same default widget and customizes it from settings or per-page script attributes. - 06
Leads & demo booking
Optional email capture, qualifying questions, and booking in the widget.
Email is never required to chat. When a visitor shares one, it becomes a structured lead with page, UTM and qualifying answers. Book a demo opens your Google Calendar appointment schedule inside the widget.
Explore - 07
Living knowledge base
Dated, owned, versioned facts with a freshness guard.
Every fact has a source, an owner and a last-verified date. Facts move from draft to published with version history and rollback, and changing one never needs a deploy.
Explore - 08
Review queue & weekly reports
See what the agent missed and fix it in one click.
The review queue collects unanswered questions, thumbs-down replies, unclaimed handoffs, blocked answers and changed source pages. A weekly report per project rolls up answer rate, handoff rate, time to claim, capture rate, CTA conversion and LLM cost.
Explore - 09
Multi-project & roles
One install, every product, each with its own agent.
One organization holds many projects. Each has its own knowledge base, persona, embed key, domains, CTAs and settings, and adding a product is configuration, not engineering.
Explore - 10
Signed-in users
Know who's asking inside your app, without spoofing.
Inside a signed-in app, call
Exploremango.identify()with a signature from your server. Visitors then get support without re-explaining who they are, and their anonymous history merges into the contact. - 11
Model routing & cost controls
Cheap models first, fallbacks, and a daily budget.
An LLM router picks the model per project with a fallback chain. Gemini Flash-Lite is the default. A daily token budget per project switches to email capture when it runs out.
Explore - 12
Greetings, A/B & cross-sell
Proactive, page-aware openers with a frequency cap.
Greeting rules open a conversation on a matching page after a delay, with up to three A/B variants per rule and a per-visitor cap. Cross-sell rules point buying intent to your other products.
Explore