Frequently asked questions

WhatsApp, Instagram, Facebook Messenger, Telegram, email read straight from your own mailbox, and an embeddable website chat widget, all from a single assistant and a single inbox.

You connect WhatsApp from the panel through Meta's WhatsApp Business platform. Business verification with Meta can take a few days, so it is best to start it early.

Messenger is tied to a Facebook Page, not to a personal profile, so you need an admin role on that Page. From the panel you can either sign in with Facebook and pick the Page, or paste a Page access token; we resolve the Page from the token and subscribe it to messages for you. Generate the token from Business Settings, System Users where you can, because a Page token minted that way does not expire.

You connect a mailbox you already own, such as info@ or support@, and the assistant reads it over IMAP and answers from your own address over SMTP. Nothing in your DNS changes and customers never see a TalkOmni address. You enter the mailbox address, the IMAP and SMTP servers with their ports, and a password; most providers require an app password rather than your normal login password. If your shared address is an alias on someone's personal mailbox, create a filter that files that mail into its own folder and point the Folder field at it, so the assistant reads only that folder. Press Test connection first: reading and sending are checked separately, so you know which half needs fixing.

Yes. TalkOmni has its own web chat widget you can customize and embed on any site with a single snippet.

TalkOmni is inbound-first. Proactive WhatsApp marketing is restricted in some regions (including Turkey), so the assistant focuses on answering, acting and handoff rather than bulk outreach.

It answers from your own uploaded documents using retrieval (RAG), so responses are grounded in your knowledge rather than generic guesses.

It replies in your customer's own language automatically, so the same assistant serves Turkish, English and many more.

Upload or replace documents in the panel at any time. New knowledge is embedded and available to the assistant right away.

Instead of guessing, it can say it doesn't know and hand the conversation off to a human agent with full context.

Yes. Beyond answering, it can take real actions against your backend, such as checking an order, looking up an invoice or booking an appointment.

Reads are open, but writes are whitelisted and validated server-side, and every action is audit-logged. The model carries intent, never prices or IDs it shouldn't.

Yes. Each project connects to your own backend through a secure, per-tenant gateway, so the assistant acts on your real data.

Yes. In addition to the built-in tools, you can define your own custom actions in the panel, wired to your backend.

Yes. Any conversation can be handed off to a human agent in the shared inbox, with the full conversation history.

No. Once a human takes over, the AI stays silent so there is no double reply, and it can auto-resume after the conversation is resolved.

Yes. Handoffs notify the right people by email and an in-app notification center, so no request waits unseen.

Yes. It runs an AI voice agent that answers inbound calls and places outbound calls in your customer's language, grounded in your knowledge.

You can buy a number in the panel or connect one you already own, then point it at a TalkOmni project.

Voice is billed by the minute, with an allowance included in your plan and top-ups available when you need more.

Yes. A visual, no-code wizard lets you build landing pages and embeddable forms, fully branded to you.

Yes. Connect your own subdomain and TalkOmni verifies ownership and issues SSL automatically.

Every submission lands in your lead pipeline, where you can track, search and export it, and it connects to your assistant.

Submissions are protected by Turnstile, per-form rate limits and origin allowlists, so bots and junk are filtered out.

It lets your team ask your own database questions in plain language, turning them into safe, read-only SQL (semantic SQL) over your real schema.

No. It is read-only by design: it can only run validated read queries over allowlisted tables, and every query is logged.

PostgreSQL, MySQL and Microsoft SQL Server, connected through a secure gateway alongside your own backend.

No. It queries your existing database in place through the gateway. Nothing is copied to us and your data never trains a model.

TalkOmni has simple monthly or annual plans. See the Pricing page for current tiers, limits and add-ons.

One AI reply is one answer the assistant writes to a customer. A conversation normally takes several replies, so a plan's reply quota is not the same as a number of conversations. How many replies a conversation needs depends on your business: an opening-hours or location question is often answered in one, while a detailed consultation runs longer. Your panel shows real usage from the first day, so the 14-day trial is the fastest way to see your own figure before choosing a plan.

Because a reply is what actually costs us: every answer runs the AI model. Pricing per conversation would mean a two-message chat and a forty-message chat cost the same, which either raises the price for everyone or forces the assistant to be cut off mid-conversation. Metering replies keeps the AI included in every plan instead of being sold as a separate module, and you only pay for answers the assistant actually wrote.

Nothing is charged automatically and no message is lost. When the quota is reached the conversation is handed to a human: messages keep arriving in your inbox, your team can answer them as usual, and the account owner is notified. You add a reply pack whenever you want the assistant to start answering again.

Add-ons are optional capabilities you can attach to a plan, such as extra voice minutes or niche action packs, billed on top of your subscription.

Yes. You can cancel anytime from your account settings; cancelling during the free trial costs nothing.

Yes. Plans can be billed monthly or annually, with annual billing offered at a discount.

Every tenant is isolated, keys are encrypted per tenant, and no query crosses between tenants. Access is scoped and audited.

No. Your documents and data are used to serve your assistant only; they are not used to train models.

We follow KVKK and international data-protection standards; see the Privacy Policy for details on how data is handled.

Scoped tokens are preferred over raw keys, whatever is stored is encrypted at rest per tenant, and secrets are never logged or returned in responses.