Features

Everything an AI agent platform should do.

Tendabot is the full loop: build a bot, train it on your business, run it on any model, talk to customers everywhere, and pay for exactly the tokens you use. Here's each piece in detail.

01 · Model freedom

42 models. 11 providers. One picker.

Every bot chooses its own brain. Run support on a fast, low-cost model and sales on a flagship — and change your mind in two clicks, with no migration and no lock-in. Prices are shown per reply before you pick.

Browse the full model catalog →

  • Claude, GPT, Gemini, Grok, DeepSeek, Mistral, Qwen, GLM, Llama — the real frontier, not one vendor
  • Per-bot selection with recommended / fast / low-cost tags and a credits-per-reply estimate
  • Switch models any time — conversations and knowledge stay put
  • Fail-safe routing: a missing provider key never takes your bot offline
02 · Live inbox

Watch the AI work. Step in whenever.

Every conversation across web, WhatsApp and email streams into one inbox in real time. The Coexistence rule is simple: when a human is driving, the AI stays silent — and hands the thread back the moment you're done.

  • Take over any chat with one click; hand back to AI just as fast
  • Replies render rich formatting — bold, lists, links — on both sides
  • Lead capture built in: name and email land on the conversation with a consent receipt
  • Assign conversations to a specific bot in multi-bot workspaces
03 · Email continuation

The conversation outlives the tab.

A visitor leaves their email and closes your site — nothing is lost. Follow up by email from the inbox; their reply threads back into the same conversation. Each conversation gets its own signed reply address, so routing is spoof-proof.

  • One-click ✉ Email follow-up on any conversation with a captured address
  • Replies land in the same thread, next to the web chat history
  • Per-bot choice: the AI answers email automatically, or a human does
  • Cryptographically signed reply addresses — foreign mail is dropped
04 · Knowledge

Trained on your business, not the internet.

Paste your docs, FAQs, policies and product details. The agent retrieves the relevant pieces for every question and answers in your voice — and you can rehearse it in the playground before a single customer sees it.

  • Per-bot knowledge — your support bot and sales bot know different things
  • Semantic retrieval picks the right passages for each question
  • A system prompt defines tone, rules and escalation behavior
  • Playground shows exactly which knowledge was used for an answer
05 · Billing engine

Metered to the token. No markup games.

1 credit = $0.01 of real model cost, passed through at cost. Every reply is measured (input and output tokens), priced from a live rate table, and written to an append-only ledger you can audit line by line.

See plans & credit pricing →

  • 300 free credits at signup; plans include monthly credits; top-ups from $25
  • Out of credits? Bots pause politely — you're never surprise-billed
  • BYOK: add your own provider key and that provider costs 0 credits
  • Live balance, 30-day burn and full ledger on the Billing page
06 · Channels & safety

One line of code. Guardrails by default.

The widget installs with a single script tag and inherits your branding. Public pages are protected from abuse out of the box, and new accounts verify their email before anything runs.

  • Website widget + shareable /try link per bot; WhatsApp-ready architecture
  • Origin allow-lists, per-IP and per-session rate limits on the public endpoint
  • Daily reply caps per visitor and per site — strangers can't drain your wallet
  • Verified-email signup, hashed single-use links, encrypted keys at rest
Ready when you are

Your first agent, answering in five minutes.