Answer like a team of twenty. Stay calm doing it.
Every customer message is read by Calmline's models: six emotions scored, intent classified, your own policies retrieved, a grounded reply drafted, and anything risky routed to a human. In about two seconds.
Pick a ticket. Watch Calmline take the first pass.
This is the real workspace. Choose a sample or paste a message from your own inbox, set tone and language, and run it. You'll see every stage of the agent, the raw model output, and the reasoning behind the draft.
Three steps from a scary inbox to a quiet one.
Calmline sits between your inbox and your team. It never sends anything you haven't approved unless you tell it to.
Bring in your messages
Forward your support email, connect WhatsApp Business, Instagram DMs or your Shopify inbox. Or just paste a message, like in the console.
The models read every one
Each ticket gets six emotion scores, an intent, an urgency level and a churn-risk number. The angry ones rise to the top of the switchboard.
Review, tweak, send
A draft that cites your policies, written in your voice, is waiting. Edit a line, switch tone, or send it as it is. Next steps go to your checklist.
Emotion mapping
Anger, frustration, anxiety, urgency, trust and satisfaction, scored 0–100 and pinned to the exact words that triggered them.
Churn risk
Spots the customers who are one bad reply away from leaving, before you've opened the ticket.
Brand-voice drafts
Learns from your best past replies. Warm, professional, concise or apologetic, at the click of a tone.
Policy-grounded
Every reply cites the exact policy lines it used, and each citation is checked against your knowledge base before you see it.
Pidgin, French, Spanish
Reply in the language your customer wrote in, including Nigerian Pidgin, without hiring for it.
Smart escalation
Chargeback threats, safety issues and big-money orders get routed to a person, with the reason attached.
A small, fast agent built for one job.
Two specialised models and a retrieval layer, wrapped in guardrails that check every draft against your own knowledge base before a person ever reads it.
4-class, on held-out real tickets from beta stores.
Across six emotions, with span-level attribution.
Triage only · triage plus full drafted reply.
English, Nigerian Pidgin, French, Spanish, reviewed by native speakers.
Whole threads, order history and your full policy doc in one pass.
of your data used to train shared models. Delete your workspace anytime.
Internal benchmarks · CalmEval v3 · n = 12,400 tickets from 340 beta stores · measured Sept 2026.
Small teams, steadier inboxes.
Averages across 340 stores in our beta, measured over their first 60 days.
"It's two of us running the shop. On Black Friday Calmline sorted 600 messages and told us which 40 actually needed us. I slept that week."
"The Pidgin replies are better than mine, honestly. Customers reply with thank-you voice notes."
"The escalation flag caught a chargeback threat I'd have skimmed past."
Priced per reply, not per panic.
Starter
For new shops answering their first customers.
- 50 AI replies a month
- 1 seat
- Triage: emotions, urgency, churn
- 4 tones, 4 languages
Growth
For busy two-to-three person teams.
- 1,000 AI replies a month
- 3 seats
- Brand voice trained on your replies
- Escalation rules and shared checklist
- Email + WhatsApp inbox
Scale
For stores with real volume and several channels.
- Unlimited AI replies
- 10 seats
- Shopify, Paystack, Instagram integrations
- Analytics: churn, CSAT, response time
- Priority support
Paid plans start with a 14-day trial. No card required.
Built by people who've worked the queue.
We started Calmline after one too many Saturday nights spent answering "where is my order?" by hand.
Ibrahim Bello
Ran support at a fast-growing Nigerian e-commerce startup, from 2 agents to 40.
Hannah Cole
Former conversational-AI engineer; built dialogue systems that handled millions of chats.
Yusuf Danjuma
Onboards every new store personally and still answers our own support inbox.
Mei Lin Tan
Designs for calm: fewer alerts, clearer queues, and screens that don't shout.
Questions shop owners ask us.
Will Calmline send replies without me?
Only if you switch on auto-send for categories you choose, like shipping-time questions. Everything else waits as a draft for you to approve.
How does it know my policies?
You give it a business context: return window, delivery times, refund rules, how you sign off. The retriever pulls the exact lines each reply relies on, and the guardrail checks every citation. If a customer asks for something outside policy, Calmline flags it instead of agreeing.
Is the Nigerian Pidgin actually good?
It's reviewed by native speakers on our team and tuned on real support conversations. You can always edit before sending, and switch any reply to English in one click.
What happens to my customers' data?
Messages are encrypted in transit and at rest, never used to train shared models, and you can delete your workspace's data at any time from settings.
What counts as a "reply"?
One drafted reply to a customer message. Regenerating the same reply in another tone doesn't count again. Triage alone is free.
Can I cancel anytime?
Yes. Plans are month to month unless you pick annual, and you can drop back to Starter whenever you like.
Get your weekends back from the inbox.
Join the waitlist for WhatsApp auto-triage and get 3 months of Growth free at launch.