By the Onsites AI team · Last updated · 5-minute read
Tawk.to is one of the internet's genuine free-forever stories: a capable chat widget at no monthly price, live on millions of storefronts, good at exactly the thing it promises — a bubble in the corner and a window where a visitor can ask a human a question. The trouble starts when support succeeds. The window fills, the questions stop being single-shot ("do you ship to Oman?") and become threads ("my order #1041 changed vessels — resend the invoice"), and suddenly the team is doing help-desk work inside chat hardware: no shared queue, no assignment, no history past the session, no CRM, and no answer to "what did we promise this customer in May?". This guide does not argue you should rip the widget out. It argues for the moment when you put a real desk behind it — and notes that with Onsites, the desk is free to three seats with the chat kept as one channel among seven, $15 a seat beyond.
Every chat-only desk recognizes these in order. History evaporates: the widget's conversations die with the session, so "what we told you last Tuesday" is folklore, and anyone who left the tab unanswered at 6 pm arrives to a thread the team cannot reopen. One human, one window: chat has no queue — two people cannot own a conversation without stepping on each other, and an absent teammate strands theirs. The channel gap: customers who will not chat (email, WhatsApp, Telegram, WeChat) fall to a personal inbox the tool cannot see — the omnichannel gap in miniature. Commercial questions with no commercial tools: "send me a quote", "why is the invoice late" end as copy-paste into documents the software cannot track. Each strain is survivable at ten chats a week and structural at fifty — which is why the upgrade conversation is about shape, not quality.
The upgrade is not a bigger chat window; it is an operating system behind it. The queue: every conversation — chat included — lands in one shared inbox with ownership, assignment and the routing rules that send invoices to finance and freight questions to the person who speaks the depot's language. Unanswered becomes unfinished: a chat nobody caught becomes a ticket nobody loses; the first-response clock finally exists because the platform measures it. The customer joins the conversation: the desk's CRM carries the account, the contact and the history — the widget's anonymous stranger becomes a record, and conversations auto-link to it. The commercial loop closes: "send me a quote" is answered inside the tool from the same editor that later bills the invoice, with the product catalog shared. None of this removes the bubble. It is why the bubble's answers get faster.
Honest answer: decide by where the conversations live, not by loyalty. If your team answers chat live, all day, from one screen, and the questions stay pre-sales — the widget can sit as a light front door while the desk runs email and the messaging channels; some desks keep both for a quarter. But notice the double answer problem: two tools means two places a customer's question can disappear, and the tool sprawl tax is real at even two products. When chat becomes a load-bearing channel — orders, complaints, "the invoice is wrong" — consolidate into the desk's own widget so the transcript, the ticket, the CRM record and the invoice share one thread. The migration cost is one snippet swap on the site and an afternoon of habit; on the metered desk it prices as seats and storage, not a new subscription line.
Make it concrete with a typical Tuesday. Before: 31 chats, two abandoned at shift change because nobody owned the window, one "resend invoice" answered by copy-pasting from a finance spreadsheet, and a WhatsApp message about container EISU-4821 noticed at 9 pm on a personal phone. After: the same 31 chats arrive with owners; the abandoned two sit as tickets flagged overdue; "resend invoice" is answered from the thread itself because invoice #1041 lives there; the WhatsApp message lands in the queue with the customer's card beside it and gets answered by the colleague covering the Shanghai morning. Same humans, same volume, none of the losses. That contrast — not a feature list — is what teams actually buy, and the desk's free tier prices it at zero until the team's third hire. The numbers that move (first response, idle threads, missed-channel rate) are all downstream of the same change: conversations that persist and belong to someone.
Seriously — an afternoon. Step one: stand up the desk (free tier takes minutes; connect email first so the inbox is alive). Step two: swap the widget snippet — the new tag replaces the old one and carries the same bubble, but transcripts now persist into the queue. Step three: invite the team; ownership appears, so yesterday's unanswered session is today's assigned task. Step four: connect the second channel — most teams pick WhatsApp or Telegram on day one — and watch a week of messages that used to hit personal phones land in the queue instead. Step five: for anything historical, export what the widget offers and forward the threads that matter; the migration guide's email-first pattern carries the rest. The only real work is cultural: reps who treat chat as a chat again learn it is now a place where order lookups happen mid-reply. That reflex takes about a week and pays forever.
The fair-loss column: a site whose support is strictly pre-sales ("what's your lead time?"), answered live by whoever is near, gains nothing by adopting a queue — Tawk.to's zero-cost bubble is the correct tool and a remarkable one at that price. A solo founder in the first month does not need a CRM yet. But every signal of success — repeat questions, orders mixed with chat, a teammate hiring — points the same direction: the widget keeps being a fine door, while the building behind it needs rooms. The complete free desk keeps the immediacy and adds the persistence: three seats, the whole product, no expiry, no per-resolution fees, AI at published credit prices. Rerun the comparison when chat stops being a corner of the work and becomes the front hall — that is the moment, and it usually arrives before the team notices.
Is Tawk.to really free?
Yes — the chat widget and its live agent window are free with paid extras. The honest limits are structural, not financial: no shared queue behind the window, conversations that end with the session, and no CRM, tickets or invoices attached to them.
What is the best Tawk.to alternative when we outgrow chat?
Onsites AI: a shared 7-channel inbox (chat as one channel), CRM with auto-linked conversations, quotes, orders and invoices from one editor — free forever on three seats, $15 per seat beyond, no per-resolution fees. Keep the bubble; add the desk behind it.
Can I keep my existing chat widget and just add a desk?
For a quarter, yes — some desks run both. But two tools means two places a question can disappear. When chat becomes load-bearing (orders, complaints, invoices), swap the snippet so transcripts persist into the shared queue.
How long does it take to add a desk behind our widget?
An afternoon: connect email, swap the widget tag so transcripts persist, invite the team for assignments, connect one more channel. The real work is a week of reps learning that chat replies can now pull the customer's order history.
When is Tawk.to the right tool even so?
Purely pre-sales sites answered live by whoever is near, and solo founders in their first month. The upgrade signals are repeat questions, orders mixed into chat and the first hire — that is when the queue, CRM and invoicing start earning their keep.