Support

Something need a human?

Support is by email. Oneop does not publish a response time — there is no SLA instrument behind one — so the fastest thing you can do is send a complete first message.

Email support@oneop.ioOpen your workspace

Worth including

  • The email you log in with
  • The URL of the screen it happened on
  • What you expected, and what happened instead

Goes somewhere else

Product · Live chat

Website chat answered by your team, with the reply already drafted.

Real-time chat from your site into a shared inbox. AI writes the first draft on request; a person edits it and presses send. And because the visitor, the deal and the invoice are the same record, your agent isn’t answering blind.

Sign-up opens when the app launches. We will email you once — no marketing list.

AI drafts. A human sends.

The AI-reply button reads the last ten turns and returns a suggested reply for the agent to edit or discard — never sent automatically.

Eight languages, two of them right-to-left

Fully translated, 7,700+ keys each. Arabic and Urdu are proper right-to-left layouts that reach into the invoice PDFs your customers receive, not just the app your agents look at.

The chat and the customer are the same record

The visitor in the thread resolves to the CRM contact and the finance customer that already exist, with their deals, tickets and invoices on the same record. Nothing syncs, because nothing is separate.

Try it here

Oneop’s own live chat runs on this site.

The website chat described on this page is the same product Oneop runs for its own customers — not a demo build. The launcher on oneop.io right now is this feature, live, running the released code.

Ask it something. Ask for a person and watch the handoff below happen to us.

Shared inbox

One inbox, five views, real presence underneath.

AllMineUnassignedMentionsVIP

From a thread, reply publicly, add an internal note, reassign, change priority or resolve — canned responses drop in with {{placeholders}} resolved live against the record. Underneath runs a real WebSocket layer: authentication, presence, typing, delivery receipts, session resume and load balancing across the team.

Both sides of every conversation are written through a single code path — visitor and bot lines persist the same way, so an agent picking up a thread sees everything that was actually said.

Inbox — Unassigned

Layla H. — website chat

“Can I get an update on order #4021?”

VIP

Unassigned visitor

Handed off from chatbot — queue position 1

Unassigned

Internal note

Not visible to the customer

Note

Handoff

When the bot hands over, it is three visible steps.

  1. The bot decides it’s done

    The visitor asked for a person, or a branch in the flow routed to the handoff block.
  2. The conversation joins the queue

    Offered to available agents against their configured capacity. No-agent-available and after-hours branches both exist.
  3. An agent takes the thread

    With the full transcript already loaded — bot turns included.

How the chatbot is built →

Where the depth goes

Two channels, and real depth on both.

A chat widget on your website, backed by email through the help desk — real SLA clocks and a full CRM behind the same record.

A website chat widget

Floating launcher and text-bubble panel, one script tag, accent colour and bottom-right position fixed in the loader.

Email, through the help desk

The platform's other customer channel — see the help desk.

Click-to-call

Hands off to your device's own dialer and logs one activity row.

Honest labeling

What is not AI here.

Some panels in the Oneop inbox look like AI and are not, so here is the honest split.

Genuinely AI, and metered

The suggested-reply button, AI triage on existing tickets, similar-ticket suggestions, and the AI draft in the ticket composer. Each is a real Google Gemini call that charges your workspace credit pool.

Labeled plainly, not dressed up

The conversation summary, sentiment label and “next best action” hint in the inbox side panel are fast client-side templates — no model call, no metering. The full business-hours SLA clock lives in the help desk.

Languages & RTL

Eight languages, including right-to-left.

Your agents work in their language and your customers read theirs — roughly eight thousand translated keys per locale, Arabic and Urdu complete right-to-left implementations.

RTL reaches the documents, the part competitors usually skip: an Arabic invoice PDF renders with an embedded Arabic typeface, a fully mirrored layout and locale-correct money and number formatting. Individual records can also carry per-language translations, so a multilingual team sees its own wording on the same record.

Arabic (RTL)GermanEnglishSpanishFrenchTurkishUrdu (RTL)Simplified Chinese

What it costs

Live chat is on every plan, including Free.

The chat itself isn't metered — only AI actions are, against a pool of 50 to 15,000 credits a month depending on plan. What scales with plan is concurrent conversations: 2 at once on Free, up to 50 on Business.

See pricing →

Control surfaces

Every AI action here runs through the same governed controls.

Prompt-injection blocking and PII redaction are on for every workspace on every plan, with no setting to turn them off — the text going into the model was typed by a stranger on the internet. Every call passes one metered choke point; at your cap you get a refusal, not an overage bill.

How Oneop governs AI →

FAQ

Common questions

Does the AI reply to customers automatically?

Not in live chat. The AI-reply button drafts a response from the recent conversation and hands it to your agent to edit and send. There is a separate auto-responder for tickets, which is opt-in and queues its drafts for human approval before anything goes out.

How many chats can we handle at once?

Two concurrent conversations on the Free plan, rising to fifty on Business. The ceiling is enforced in the real-time layer, and a visitor arriving over it is told to try again rather than left on a chat that never connects.

Is the inbox actually real-time?

Presence, typing indicators and delivery receipts are pushed over a WebSocket the moment they happen. The thread list itself refreshes on a fast, short poll.

Does the inbox summarise conversations with AI?

The summary, sentiment and next-best-action panels are fast client-side templates, not metered model calls. The genuinely AI features are the suggested reply, ticket triage and similar-ticket suggestions.

Which languages does the agent interface support?

Eight, fully translated: Arabic, German, English, Spanish, French, Turkish, Urdu and Simplified Chinese. Arabic and Urdu are full right-to-left, including in generated invoice PDFs.

Get started

Answer chats with the whole customer on screen.

Free plan, two seats, two concurrent chats, no credit card.

Sign-up opens when the app launches. We will email you once — no marketing list.