Turn years of experience, research, and accumulated knowledge into an AI assistant that shares your expertise 24/7.
"Yes — annual plans include 2 months free, applied automatically at renewal."
Source: pricing.pdf §3Upload manuals, policies, FAQs, and support articles — or audio and video recordings, which are transcribed automatically. Or just add your website, or link your Dropbox folders: pages and files are read in and kept up to date. They become the source of truth your agent draws every answer from.
Pick a model, connect it to a knowledge base, and shape how it talks. Each agent holds a conversation the way your team would.
Clients can send voice messages in Telegram — the assistant understands them and can answer with a voice message of its own, in 18 languages, in a natural voice you choose and preview.
Group knowledge bases, agents, and channels into projects. Usage-based credits scale with how much your agents actually work.
// channels
Connect the same agent to the sources your customers already use to reach you.
Paste one script tag into your site. A chat bubble appears in the corner, answering from your knowledge base.
Connect your assistant's token once. When it gets a message — typed or a voice message — the AI assistant finds the relevant information and replies, in text, by voice, or both.
Link a WhatsApp Business number through Meta's Cloud API. Customers message the number they already have saved.
Call the same chat endpoint directly from your own app or backend, for integrations the built-in channels don't cover.
// pricing
Simple, usage-based pricing. Credits scale with how much your agents actually work, so you only pay for the answers they deliver.
Try AOVA with your own documents.
For experts and small teams going live.
For busy channels and several agents.
For high-volume support at scale.
Need more? Top up anytime: 1,000 credits for €15. Top-up credits never expire.
Speech to text (voice messages, audio and video recordings): 8.8 credits per minute. Voice replies: from 0.8 credits per minute.
* Based on a typical answer on Claude Haiku 4.5 (≈ 2 credits). Amazon Nova answers use ≈ 1 credit, Claude Sonnet ≈ 5. Prices in EUR.
// ai models
Choose the AI model that works best for your data, use case, and desired response style.
Claude Haiku 4.5 · Claude Sonnet 4.6
A strong choice for complex knowledge bases, detailed answers, analysis, and nuanced conversations.
Nova Micro · Nova 2 Lite
Fast and cost-efficient, from lightweight everyday answers to more capable reasoning — a great fit for scalable AI assistants.
GPT-6.1 Sol · GPT-6 Luna
Versatile models for everything from quick, lightweight replies to advanced reasoning and in-depth answers.
Gemini 3.8 Flash · Gemini 3.1 Flash-Lite
Fast, efficient responses at scale — well suited to high-volume assistants that need quick answers.
Grok 4.7
A direct, conversational style for assistants that keep answers short and to the point.
You can use your existing knowledge base, documents, guides, articles, educational materials, FAQs, internal documentation, and other sources of information — including audio and video recordings such as calls, webinars, and lectures, the pages of your website, and folders in your Dropbox. Your knowledge becomes the foundation of your AI assistant.
Yes. Paste your website address and choose to read one page, the whole site, or the pages in its sitemap. AOVA reads the public pages into your knowledge base and can read them again weekly or daily, so when you change a price on your site, your assistant knows. Pages behind a login, or drawn only by JavaScript, can’t be read — upload those as files. Reading pages costs no credits; each plan includes a number of pages (Free 20, Pro 500, Business 5,000, Enterprise 20,000).
Yes. Click Add knowledge → Dropbox, allow read-only access once, and tick the folders (or files) your assistant should know. When you edit, add or delete a file in Dropbox, your assistant follows within minutes. Documents are read right away; audio and video are listed with their length and price and only transcribed when you approve them. If a file is deleted in Dropbox, its metadata and transcript are kept for 30 days and come back automatically if the file does. AOVA never changes your Dropbox. Each file uses one page of your plan, shared with websites.
Yes. Upload MP3, WAV, M4A, FLAC, OGG, AMR, WEBM, or MP4 files (up to 1 GB and 4 hours each) alongside your documents. Speech is transcribed to text automatically — in English, Russian, and other major languages — and your assistant answers from it like any other document. Transcription is billed in credits per minute of recording, and you see the estimate before you upload.
Yes, in Telegram. Switch on "Understand voice messages" for a Telegram channel and clients can simply speak: each voice message (up to 5 minutes) is turned into text and answered like a typed question. The assistant can also answer by voice — as text, as a voice message, or both — always, or only when the client spoke. You pick the voice and its quality (Standard, Natural or Premium) and can listen to it first. Understanding voice costs 8.8 credits per minute of voice message; voice replies cost from 0.8 credits per minute, depending on the quality. WhatsApp and the website widget answer text messages.
Text answers work in any language your clients write in. Voice works in 18 languages: English, Russian, German, French, Spanish, Italian, Portuguese, Dutch, Polish, Swedish, Danish, Finnish, Romanian, Arabic, Hindi, Japanese, Korean and Chinese. Choose up to 5 for an agent — the assistant recognises which one each client speaks and answers in it. Audio and video recordings in a knowledge base are transcribed in English, Russian, Ukrainian, German, French, Spanish, Italian, Portuguese, Polish and Turkish.
Yes. Open the person's conversation and choose Block contact. From then on their messages never reach your assistant and cost you no credits, but they are still saved so you can see what was sent. For each channel you decide what a blocked person sees: nothing at all, or a short message of your own (sent at most once a day). You can unblock them at any time.
No. You don't need to build or train a model from scratch. Simply provide your knowledge base and choose the AI model that best fits your needs.
You can choose from leading AI models, including Anthropic Claude, OpenAI GPT, Google Gemini, and Amazon Nova. Different models can offer different advantages in reasoning, speed, cost, and other capabilities.
Yes. You can switch between available models as your needs change, without having to rebuild your knowledge base from scratch.
No advanced technical knowledge is required. The platform is designed to make creating a knowledge-based AI assistant simple, even if you're not an AI developer.
A general AI chatbot is designed to answer questions using its broad capabilities. Your knowledge-based assistant is designed around your information — making it possible to turn your accumulated expertise and documentation into an interactive source of knowledge.
It's especially useful for experts, consultants, educators, companies, and teams that have accumulated a significant amount of valuable knowledge and want to make it easier to access, share, and scale.