Dvaarik is self-serve SaaS today: there is no setup fee and no minimum commitment. You log in, add your services, staff, hours and knowledge, connect channels, and pay only for what your agents actually do. | What | Price | Notes | |---|---|---| | Setup fee | ₹0 | Self-serve — configure your business from the dashboard | | AI voice calls | from ₹2/min | Simple per-minute billing on whole minutes; higher grades up to ₹5/min | | AI chat | ₹2 per conversation | 24-hour window | | Phone number + 1 concurrent line | ₹1,500 per 30 days | Includes up to 3 numbers; debited from credits | | Social Plan — WhatsApp + Instagram agents + campaigns | ₹1,500 per 30 days | Only if you use those channels | | WhatsApp message charges | ₹0 markup | Meta bills your own WhatsApp Business account directly | Credits are prepaid and never expire; at zero balance agents pause gracefully, so there are no surprise bills. Agencies and developers can sell the same platform under their own brand through the white-label partner programme, and set their own client pricing.
What does an AI receptionist actually cost a restaurant per month in India?
Dvaarik is self-serve SaaS today: there is no setup fee and no minimum commitment. You log in, add your services, staff, hours and knowledge, connect channels, and pay only for what your agents actually do.
| What | Price | Notes |
|---|---|---|
| AI voice calls | from ₹2/min | Simple per-minute billing on whole minutes; higher grades up to ₹5/min |
| AI chat | ₹2 per conversation | 24-hour window |
| Phone number + 1 concurrent line | ₹1,500 per 30 days | Includes up to 3 numbers; debited from credits |
| Social Plan — WhatsApp + Instagram agents + campaigns | ₹1,500 per 30 days | Only if you use those channels |
| WhatsApp message charges | ₹0 markup | Meta bills your own WhatsApp Business account directly |
The comparison above states the published row-by-row differences; read each row with its source, date, and qualifying notes.
Credits are prepaid and never expire; at zero balance agents pause gracefully, so there are no surprise bills. Agencies and developers can sell the same platform under their own brand through the white-label partner programme, and set their own client pricing.
What do AI receptionist vendors in India charge a restaurant per minute, and what does it take to get that rate?
Headline per-minute rates in this market are mostly entry tickets to a sales conversation. The number below that matters is not the rate — it is the second column. Every figure was read on the vendor's own page on 21 Jul 2026 and is stored with the sentence it came from in our India AI voice agent pricing index.
| Vendor | Published rate | What it takes to get that rate | Checked |
|---|---|---|---|
| Social Plan (optional — WhatsApp + Instagram agents + campaigns) | ₹1,500 per 30 days | Only if you use those channels | |
| Scalify Labs | ₹0.40/min | Self-labelled estimate — their own note reads "Rates are estimates… Always get a formal quote". Hindi + English only. Minutes are billed on top of a plan from ₹15,000/mo, plus ₹15,000–₹40,000 one-time setup, a 10,000 min/month minimum and a 3-month commitment | 21 Jul 2026 |
| Agni by Ravan.ai | ₹2/min headline | Their rate card puts ₹2/min on quote-only Enterprise at 10,000+ minutes. The buyable entry is ₹2,999/mo with 300 minutes and ₹8/min overage; Starter is Hindi + English only | 21 Jul 2026 |
| HuskyVoice.AI | ₹4/min | Published verbatim as "Enterprise volume pricing as low as ₹4/min" — a volume floor, not an entry rate. The entry plan is ₹1,999/mo including 100 voice credits at 2 credits per minute | 21 Jul 2026 |
| ConnectAI | ₹4/min (8 credits) | A subscription is required to unlock the receptionist: ₹800/mo standalone or ₹499/mo as an add-on to their ₹2,499/mo suite, plus a one-time ₹1,000 clinic onboarding fee for new clinics. Product is clinic-scoped | 21 Jul 2026 |
| VaniAgent | from ₹4.5/min (₹4.5–5.5) | Metered credits, but a monthly plan is the entry point either way: the metered tier is ₹4,999/mo, and the flat unlimited tier is ₹21,999/mo with a 3-month minimum on the unlimited plans | 21 Jul 2026 |
| Trikon | ₹5/min | Flat pay-as-you-go, no setup fee, no voice plan — but telephony and DID numbers are "billed separately by your carrier", so a phone bill sits on top of the ₹5 | 21 Jul 2026 |
| Bolna AI | 6.00¢/min, shown as ₹5.52 by their own conversion | Prepaid credits, no voice subscription. Their docs state total cost = voice processing + telephony + "a flat per-minute fee charged by Bolna" whose amount is not published, so ₹5.52 is not established as all-in | 21 Jul 2026 |
| Botsense | ₹9/min Starter, ₹7/min Growth, ₹5/min Enterprise | Rates are shown openly and the page says "no monthly commitment", but there is no self-serve checkout — every CTA routes to sales, and Enterprise is a custom quote | 21 Jul 2026 |
| MyOperator | No ₹/min published | The standalone AI voice agent is ₹10,000/mo with 2,000 minutes included; overage is published as ₹8 per conversation, not per minute. All plans are billed annually, and dedicated onboarding is ₹20,000 one-time | 21 Jul 2026 |
| SquadStack | Not published | No rupee figure appears anywhere on their site. The rate is shared after a discovery call, a one-time per-use-case setup fee applies with no amount stated, and the entry tier is a 90-day commitment | 21 Jul 2026 |
The comparison above states the published row-by-row differences; read each row with its source, date, and qualifying notes.
We are not claiming to be the cheapest, and you should distrust anyone in this market who does. The claim we will defend is about terms, not the digit: ₹2/min applies to your first minute, at any volume, with nothing to sign.
One methodology note, because it changes how you should read every other comparison page you find. Several Indian vendors publish tables of *rivals'* rates. Trikon's page, for example, lists Bolna at ₹8–11/min — Bolna's own pricing page says 6.00¢, which Bolna itself renders as ₹5.52. We do not source a single number in the table above from a competitor's comparison page, only from each vendor's own site, and we date-stamp it because several of these rates moved within weeks.
Which restaurant calls are actually worth ₹2 a minute?
Not all of them, and pretending otherwise is how these projects get abandoned in month two.
A two-minute call bills ₹4. Sort your inbound by what a missed one costs you:
- Party, banquet and bulk enquiries. One captured enquiry is worth your per-head price times the guest count — for a 40-guest function that is tens of thousands, and you know that number better than we do. Today Dvaarik is self-serve with no setup fee and no required plan — usage is prepaid credits, voice from ₹2/min with whole-minute billing. This is the calculation that matters most for a restaurant with a function hall.
- Table reservations, especially weekend evenings. A missed booking call at 8pm is a table that goes to whoever picked up.
- Direct delivery and takeaway orders. Every direct order is an order that does not carry your aggregator's commission — whatever that percentage is for you, it is not zero.
- Delivery-zone and "do you deliver to my pincode" questions. Under a minute each, so one billed minute and ₹2 — and the wrong answer costs you the order.
- Menu, price, timing, address, parking. Genuinely low value per call, and genuinely high volume. This is what stops staff from answering the calls above.
The agent's job is to clear the last category instantly and route the first two to a human or into a structured capture. If your phone only ever rings for the last category, be honest with yourself and read the last section of this page.
Cloud kitchens: what changes when there is no dining room
A cloud kitchen's phone is a different instrument. There is no host stand, no reservation book, and the calls that come in are order status, delivery-zone serviceability, item availability, a repeat customer trying to order directly, and complaints.
The economics also differ. Cloud-kitchen calls are short — "is my order out" and "do you deliver to Kondapur" are one- and two-minute conversations. Billing is whole minutes, so a 70-second call bills two minutes: short calls do not buy you a fraction of a rupee, and you should plan for the rounding rather than against it. Forty calls a day, each billing two minutes, is 2,400 minutes and ₹4,800 a month. Volume can be high without the bill running away, but it is the call count, not the seconds, that drives it.
What we build first for a cloud kitchen, in order:
- A direct-order line that actually gets answered. The number on your packaging, your Instagram bio and your WhatsApp catalogue. The agent answers 24/7, takes the customer's details and pushes the conversation to WhatsApp where the items are confirmed in writing.
- Pincode and zone serviceability. A rule set, not a guess. The agent tells the caller yes or no, and the minimum order value for that zone if you have one.
- Item availability with an owner-controlled pause. When a dish is 86'd or the kitchen is drowning, you flip the agent into enquiry-only mode from your phone. It stops taking orders and starts taking callbacks.
- Order-status calls off your staff's plate. These are pure interruption cost. Handing them to the agent is usually where a kitchen manager first believes in the thing.
What we deliberately do not build on day one is full spoken ordering with customisations. See the FAQ below for why.
Party, banquet and bulk enquiries: the calls that pay for the whole month
If your restaurant does functions, this is the workflow to build first, and it is where the ₹2/min bill becomes rounding error.
A party enquiry is a long call — three to five minutes — and it is negotiated in the caller's comfort language. The agent's job is not to close it. It is to capture it completely so a human can close it: date, time, guest count, veg and non-veg split, hall or section preferred, budget per head or total, whether they need decor and DJ, and a callback window.
That capture lands as a filled lead row and a WhatsApp summary to the manager. If nobody calls back within your stated window, the Leads agent can call the enquirer back — and to be exact about what that means, outbound calling is only ever to people who contacted you or gave you their number for this purpose. Never a bought list. Never outside permitted calling hours. The agent identifies your restaurant at the start of the call.
For deposits, the Payments agent can send a UPI payment link inside the same conversation, and Dvaarik takes no commission on what you collect. The link opens the caller's UPI app with your payee name and the amount already filled in, so they approve with their PIN and the money reaches your bank account in seconds. You set the deposit amount and the rule for non-payment — hold the slot two hours then release, or escalate to the manager. See how the link flow works on the payments page, and how slot-locking works on the booking page.
Reservations, no-shows and the WhatsApp trail after the call
Restaurants run on WhatsApp memory. The call ends, and the useful part is what arrives afterwards: the menu PDF, the location pin, the booking time in writing, the advance link, the party package.
The pattern that works is simple and cheap. The voice call takes the booking. A WhatsApp message confirms it in writing — that is ₹2 on our side plus Meta's own conversation charge billed to your account. A reminder goes out a few hours before the slot with a one-tap confirm or release. A released table goes back into the book instead of standing empty at 8:30pm on a Saturday.
In our builds the reminder is the cheapest step and the one owners are most likely to leave out — and it is the one that puts covers back on the book. It costs ₹2 a conversation. A released and refilled two-top on a weekend evening is worth many multiples of that.
For demand that starts on Instagram — a reel, a story, a comment saying MENU — the Instagram agent replies in DM and moves the thread to WhatsApp. Today Dvaarik is self-serve with no setup fee and no required plan — usage is prepaid credits, voice from ₹2/min with whole-minute billing.
Rush-hour guardrails: what the agent must not be allowed to do
Most restaurant AI failures are not accuracy failures. They are permission failures — the agent was allowed to promise something the kitchen could not deliver.
The rules we hard-code on every restaurant build:
- Never assert table availability unless it is connected to a real book or a real rule set. "Let me take your details and the manager will confirm within ten minutes" is a better answer than a confident wrong one.
- Never invent stock. If it is not on the availability list, the agent offers alternatives instead of guessing.
- One-tap pause, owner-controlled. Kitchen overloaded, chef down, power cut — you switch the agent to enquiry-only from your phone and it stops committing to anything.
- Escalation rules the owner wrote. Complaints, allergy questions and anything involving a refund go to a human, immediately, with the transcript.
- No promises about delivery time beyond the window you told us to quote.
The subject described in this section: These are boring and they are the reason the agent survives past its first Saturday night.
Telugu start, English finish — the language your callers actually use
Mixed-language calls are the normal case in an Indian restaurant, not the edge case. A caller opens in Telugu, gives the guest count in English, negotiates the per-head budget back in Telugu, and confirms the date in English.
The agent handles 22 scheduled Indian languages plus code-mixed speech — Hinglish, Tenglish, Tanglish — and follows the caller when they switch mid-sentence instead of dragging them back into one language. That matters most on party enquiries, where the person calling for a 40-guest function will not negotiate in their second language.
You can fix the greeting language per outlet — a Hyderabad branch opens in Telugu, a Pune branch in Marathi, both fall back to English on request. All of it sits at the ₹2/min Bharat Standard rate; there is no regional-language surcharge on our side, which is worth checking against whichever vendor you compare us with, because language gating by tier is common in this market. More on the multilingual page.
When an AI receptionist is a bad buy for a restaurant
Today Dvaarik is self-serve with no setup fee and no required plan — usage is prepaid credits, voice from ₹2/min with whole-minute billing. That has to be earned back in calls that carry revenue.
The subject described in this section: It does not pay for itself if:
- Nearly all your volume arrives on aggregator tablets and walk-ins, and the phone rings four or five times a day.
- You are a counter-service outlet with no reservations, no functions and no direct delivery. A WhatsApp catalogue will serve you better and cost you almost nothing.
- Nobody at the restaurant can sit with us for an hour to approve the menu answers, the delivery zones and the escalation wording. An unapproved script should not take real customer calls.
The subject described in this section: It pays back quickly if you have party or bulk enquiries worth thousands each, direct delivery you would rather not hand to an aggregator, weekend reservation pressure, or more than one outlet sharing a number.
If you are unsure, tell us roughly how many calls you miss between 12–3pm and 7–11pm and we will say plainly whether it is worth it — including when it is not. Full offer on the pricing page, and how restaurants use it on the restaurants page.
Frequently asked questions
What will an AI receptionist actually cost my restaurant in month one?
Dvaarik is self-serve SaaS today: there is no setup fee and no minimum commitment. You log in, add your services, staff, hours and knowledge, connect channels, and pay only for what your agents actually do.
| What | Price | Notes |
|---|---|---|
| AI voice calls | from ₹2/min | Simple per-minute billing on whole minutes; higher grades up to ₹5/min |
| AI chat | ₹2 per conversation | 24-hour window |
| Phone number + 1 concurrent line | ₹1,500 per 30 days | Includes up to 3 numbers; debited from credits |
| Social Plan — WhatsApp + Instagram agents + campaigns | ₹1,500 per 30 days | Only if you use those channels |
| WhatsApp message charges | ₹0 markup | Meta bills your own WhatsApp Business account directly |
The comparison above states the published row-by-row differences; read each row with its source, date, and qualifying notes.
Credits are prepaid and never expire; at zero balance agents pause gracefully, so there are no surprise bills. Agencies and developers can sell the same platform under their own brand through the white-label partner programme, and set their own client pricing.
Can it take a full food delivery order on the phone, with customisations?
We would not make full voice ordering your first workflow, and it is fair to say so. Spoken orders carry variants, spice levels, substitutions, jain/no-onion-garlic requests, and address confirmation — every one of those is a place where a wrong item reaches the kitchen and you eat the cost. What works reliably from day one is enquiry capture: the agent answers menu, price, timing, and delivery-zone questions, takes the customer's details, and sends the menu plus a WhatsApp thread where the order can be confirmed in writing before anything hits the kitchen. Once that is stable and you trust the transcripts, we can extend it toward ordering for your repeat items.
My staff pick up the phone when they can. Can the AI only answer the calls we miss?
Yes — that is usually the cheapest way to start. You keep your existing restaurant number and set busy/no-answer forwarding on it, so the AI only picks up when the line is engaged or nobody reaches the handset in time, which is the situation you already know from a dinner rush. Nothing changes on your printed cards, Google listing, or aggregator profiles. If you would rather run a separate line for bookings and party enquiries, we can supply a number at ₹1,500 per 30 days, auto-debited from your credits. Either way you get the caller's number and a summary on WhatsApp, so a missed call stops being a lost table.
Can it collect an advance for a party or bulk booking?
Yes — the Payments agent can send a payment link inside the conversation, and Dvaarik takes no commission on what you collect. In practice the agent captures date, time, guest count, veg/non-veg split, and budget, then sends a WhatsApp link for the deposit amount you set (the same upi://pay link you can generate by hand). You also set the rule for what happens if it goes unpaid — hold the slot for two hours, then release it, or escalate to the manager. Read how the link flow works on the payments page, and how slot-locking and reminders work on the booking page.
Customers here start in Telugu and finish in English. Will it get confused?
No — mixed-language calls are the normal case in Indian restaurants, not the exception. The agent handles 22 scheduled Indian languages plus Hinglish, Tenglish, and Tanglish, and follows the caller when they switch mid-sentence rather than forcing them back into one language. For a restaurant that matters most on party enquiries, where the person calling for a 40-guest function will negotiate in their comfort language and give you the guest count in English. You can also fix the greeting language per outlet — a Hyderabad branch can open in Telugu while another opens in Hindi. More on the multilingual page.
When is this not worth it for a restaurant?
If almost all your volume arrives through aggregator tablets and walk-ins, and the phone rings only a few times a day, this will not pay for itself. Today Dvaarik is self-serve with no setup fee and no required plan — usage is prepaid credits, voice from ₹2/min with whole-minute billing. It pays back when you have party and bulk enquiries worth thousands each, direct delivery you do not want to hand to an aggregator, table reservations, or more than one outlet sharing a number. If you are unsure, tell us roughly how many calls you miss between 12–3pm and 7–11pm and we will say plainly whether it is worth it. See how restaurants use it on the restaurants page.
Restaurant AI reception is not about replacing hospitality. It is about catching the calls, DMs, and WhatsApp messages staff cannot handle during rush hour.
Send us your menu and the calls you miss most. We will map the first restaurant AI workflow.
LoginTagged

Written by
Rohith Sriramula
Founder & CEO, Dvaarik AI
A laid-off engineer who went all in on Dvaarik AI. He builds the platform and product workflows from hands-on work with Indian businesses, not theory.