Everything Tendabot does.

Claude, GPT, Gemini, Llama, Grok, Mistral, DeepSeek, Qwen, Groq, Cerebras and Fireworks. Switch a bot's model any time — no migration, no re-training. See the full list with live per-reply costs on the models page.

Bots run with always-on guardrails: answer from your knowledge, admit uncertainty, and offer a person rather than invent a price, a policy or an address. Grounded facts beat guesswork.

Write in German, get German. Write in English, get English — including refusals and policy answers, not just the easy questions.

Answers appear word by word as the model writes them, so nobody watches a spinner.

Replies can carry numbered source cards showing which article or page the answer came from — clickable, so a visitor can check for themselves.

For stores: the bot reasons about fit, compatibility and use case, but only ever recommends from your real catalogue.

Paste text, or upload PDF, TXT, Markdown and CSV files up to 10 MB. Documents are split into chunks and the most relevant pieces are retrieved per question.

Give it your URL. Tendabot discovers your pages, you pick which to include, and each becomes citable knowledge with a link back to the source. Re-sync any time.

Write once: published as a public help centre for SEO and taught to the bot, so the same answer serves readers and chat.

Every save snapshots the bot's instructions and settings. Restore any earlier version in one click.

Visitors can tap “Talk to a human” any time, and the AI offers it by itself when a question is sensitive or beyond its knowledge. The thread flips to your team, the AI goes quiet, and a person replies in the same window.

No one watching the inbox? Switch it off and the AI is instructed never to promise a person it can't deliver. A real off switch, not a hidden button.

Every conversation across every channel in one place, with take-over and hand-back-to-AI, conversation stages (open / pending / resolved), and full timestamps.

Per bot, pick exactly which team members are emailed on a handoff or a customer request — the person who answers chats often isn't the account owner.

Alerts reach your phone even with the console closed.

Out of hours, visitors are told when your team is next around instead of waiting on a reply that won't come until morning.

Connect Shopify or WooCommerce and the bot answers “where is my order” from your real store — status, tracking, delivery.

An order number alone is never enough. The visitor must also confirm the checkout email, the phone number, or the delivery postcode — so nobody can walk order numbers to read other people's details.

Recommendations come from your live catalogue with real prices and links, never invented products.

Cancellations, address changes, refunds, callbacks — the bot gathers the details, files the request to your inbox, and tells the customer a person will confirm. It never claims the action is already done.

One script tag on any site — WordPress, Shopify, Webflow, Wix, Squarespace or plain HTML. Plus a shareable /try link per bot with no install at all.

Connect a WhatsApp number so the same bot answers there too.

A visitor who leaves an email can be replied to after they close the tab — and their reply threads back into your inbox as the same conversation.

Visitors can send themselves a copy of the conversation. With your own mail server connected it arrives from your domain, and replies come back to you.

File, image and video uploads, voice notes (transcribed so the AI answers what was said), emoji and GIFs, typing indicators, and 👍/👎 on replies.

Optional pre-chat form for name, email and phone with your own consent text — so you get the lead even if they don't buy.

Your plan sets how many AI replies you get; credits cover what those replies cost at the provider's real rate, with no markup on tokens. Full ledger in Billing.

Busy month? Add +1,000 for $39 or +5,000 for $149. They never expire, and your included allowance is always used first.

A reply sent by a person on your team never counts against your allowance and never spends credits.

Paste your own provider key and that provider's replies cost zero Tendabot credits — they bill you directly. Keys are encrypted and validated on save.

Card not an option? We can invoice you and manage your plan directly. Ask us.

Everyone who ever talked to your bot — name, email, phone, first contact, activity — with editable profiles and notes.

Conversations, AI replies, spend per bot, cost per reply, model breakdown, busiest hours and customer satisfaction ratings.

Keyword rules that answer or route before the AI runs, auto-resolve for idle threads, and a configurable message for when something goes wrong.

Push conversation events to your own systems — CRM, Zapier, anything that accepts a POST.

A built-in MCP server: create bots, edit instructions, add knowledge, read conversations and reply — from your AI assistant.

20% commission on every payment from workspaces you refer, tracked automatically.

Colours, position, launcher icon, title, welcome message, starter suggestions, your logo and avatar, and real team faces — all without touching the embed code.

Hide the “Powered by Tendabot” badge entirely.

Origin allow-lists, per-IP and per-session rate limits, and daily reply caps per visitor and per site — so strangers can't drain your wallet.

Spend fails closed. When credits or replies run out the bot pauses politely and routes to your team; you are never billed past what you bought.

Encrypted keys and tokens at rest, verified-email signup, hashed single-use links, and per-workspace isolation on every read and write.