🧠
It answers with your information, not with generic filler.
Import your PDFs (product sheets, FAQs, documentation), scrape your website pages, write your own instructions. Native RAG splits and vectorises everything: at each question, only the relevant excerpts reach the model. Precise answers even with 500 PDFs, where classic chatbots break down after a few dozen pages.
- ✓PDFs up to 10 MB per file, URLs and free text
- ✓Your whole site explored: you tick the pages the bot should know (up to 200)
- ✓From Essentiel, pages re-read every week and updated when they change
- ✓Unlimited semantic base on Business
⚡
The engine that fits your use case, switchable in one click.
GPT-6 Luna, GPT-5.6 Luna and GPT-6.1 Sol (OpenAI), Claude Haiku 4.5 and Sonnet 5.5 (Anthropic), Gemini 3.5 Flash-Lite and 3.8 Flash (Google), Mistral Small and Large (Mistral AI, France). Switch from the dashboard without reconfiguring: your bot keeps its knowledge, personality and history. Need the whole chain in Europe? Turn on the “Data 100% in Europe” setting: answers, knowledge search and dictation then run on Mistral AI in Europe.
- ✓GPT-6 Luna, the recommended default: fast, reliable and the cheapest
- ✓The five fast models on every plan, the four more powerful ones on Pro and Business
- ✓No per-model surcharge: everything sits in your quota
🎯
Your bot doesn't just chat: it collects.
Email request before or during the conversation, customisable form with the fields you choose, an email for each new contact (right away or once a day), one-click CSV export. When a visitor wants a meeting, the bot offers your Calendly, Cal.com, Doctolib or Planity link as a real button. Every conversation can end as a qualified prospect in your inbox, with the full context of the discussion.
- ✓Form built into the chat window, no redirect
- ✓Email for each new contact, with the conversation if you want it
- ✓Booking link shown as a button in the chat
- ✓CSV export for your CRM
🎨
Your colours, your avatar, your bot's name.
Primary colour, position, bubble size, five launcher styles (bubble, message, quick replies, pill, question bar), welcome message, avatar and name: everything is set from the dashboard with live preview. Voice dictation and thumbs under answers are one switch away.
- ✓Live preview, no code
- ✓Tone, formality, forbidden topics, hand-off by email
- ✓Causerie badge removable from Pro, not a paid add-on
📊
You finally see what your visitors actually ask.
Every conversation is readable in full and filterable (active, resolved, leads, archived), with usage statistics: conversation volume, leads generated, messages, response time. It is often the most valuable discovery of the first weeks: the real questions your customers have.
- ✓Full conversation history in real time
- ✓Statistics per bot and per period
- ✓Spot unanswered questions to enrich your knowledge base
- ✓Answers to improve: visitors rate replies with a thumb, you write the right answer
🚧
Free-form instructions, counted in your plan's volume.
Define precisely who your bot is, what it must do, and what it must never say. Forbidden topics (detailed pricing, for instance), a hand-off to your team by email (conversation included) when the question goes beyond its scope, or a live takeover from the dashboard, conversation goal (capture a lead, book a meeting, qualify a quote). This is what separates a generic bot from a real digital employee.
- ✓Free-form instructions, no separate cap
- ✓Forbidden topics, hand-off by email or live takeover
- ✓One goal per bot: lead, booking, support, sales
🔒
Compliance as a default, not as a paid option.
Data hosted in the European Union, encrypted at rest and in transit, never used to train AI models. Conversations belong to the bot owner: readable, exportable and deletable at any time, kept until you delete them or for a period you choose (30, 90 or 180 days, 1 or 2 years). Allowed websites keep a copied bot from showing anywhere else, and a fallback model takes over if the main one refuses or does not answer within 25 seconds. With the “Data 100% in Europe” setting, every AI step (answer, knowledge search, fallback model, dictation) runs on Mistral AI's European endpoint, and stops rather than leave Europe if it does not respond. Without it, knowledge search goes through OpenAI.
- ✓EU hosting, encryption at rest and in transit
- ✓No training on your content, contractually
- ✓One switch per bot keeps every AI step in Europe
🔌
One snippet before your closing body tag.
WordPress (official plugin on wordpress.org), Shopify, Webflow, Wix, Squarespace, Next.js or plain HTML: the same script works everywhere, no heavy plugin, no developer. Loads deferred, after your content, and weighs about 77 KB transferred (measured on October 26, 2026).
- ✓Official WordPress plugin for a two-click install
- ✓Compatible with every CMS and framework
- ✓Deferred loading: your page speed stays yours
Every feature is included in our plans, with no hidden add-on. See which plan has what →
