Experience Lion CRM — The King of WhatsApp CRM 7 Days Free Limited Access
→ Request Access

WhatsApp Business API: What It Is in 2026

The WhatsApp Business API explained as it works in August 2026 — access rules, three live meters, India rates, and the 1 October change.

Most explainers on the WhatsApp Business API describe a product that stopped existing.

They tell you Meta charges per 24-hour conversation. It does not, and has not since 1 July 2025. They tell you the first 1,000 conversations each month are free. That allowance is gone. They tell you replying to a customer inside the 24-hour window costs nothing. That is true today and stops being true on 1 October 2026. They tell you the API is a piece of software you buy from a provider. It is closer to a licence you are granted, metered by Meta, and it can be narrowed while you sleep.

None of that is the writer’s fault. Meta has changed the commercial terms of this platform four times in two years, and the last change landed two weeks ago: since 1 August 2026, replies produced by Meta’s own AI agent are billed by token, not by message. That is a different unit of account from everything else on the platform, and it makes one kind of message roughly thirty times more expensive than another.

So this guide does not open with a definition. It opens with the honest shape of the thing: what the API grants you, what it withholds, what each of its meters charges in India this month, and how to tell whether you need it at all. Most businesses searching this term do not, and the pages that would tell them so are all published by companies that sell it.

In this guide

What the WhatsApp Business API actually is

The WhatsApp Business API is a programmatic channel into WhatsApp. It has no interface of its own. There is no app to install, no inbox to open, no screen where messages appear. It is a set of endpoints that accept a message and a recipient and hand them to WhatsApp’s delivery network, plus a stream of events that tells your software what happened afterwards.

That absence of an interface is the single most consequential fact about it, and it is usually mentioned in passing. Everything a person would want — an inbox, assignment, notes, contact history, reporting, a bot builder — is built by somebody else on top of the API. When you pay a platform, that layer is most of what you are paying for. Meta charges separately, per message, underneath.

The second consequential fact: the API is not tied to a device. The WhatsApp app and the free WhatsApp Business app both run on a phone, and a browser extension over WhatsApp Web is still that phone’s session in a different window. The API runs on a server. It fires at 3am, it fires while your team is on leave, it fires when a payment gateway calls a webhook, and it keeps firing whether or not anyone is holding a handset. Every genuine capability difference traces back to that one architectural change.

The third: it is per-number, not per-company. The API attaches to a specific phone number inside a WhatsApp Business Account, and that number cannot simultaneously be logged into the WhatsApp app or the WhatsApp Business app. Moving a working number onto the API means the phone stops receiving on it. This surprises more businesses than anything else in setup, and it is why most start with a second number.

Five things the explainers still get wrong

If you have read three articles about this API in the last week, you have probably absorbed at least two claims that are no longer true. Here is the current state, with the date each thing changed.

Common claim Status What is true now
You are billed per 24-hour conversation Retired 1 July 2025 You are billed per message, by category
1,000 free conversations a month Retired 1 July 2025 No free monthly allowance
Replies inside the 24-hour window are free True until 1 October 2026 Service messages become billable that day
Utility templates inside the window are free True until 1 October 2026 They lose in-window free status that day
Everything is priced per message Incomplete since 1 August 2026 Meta’s AI agent replies are billed per token
You need a Meta partner to use it Not required Cloud API is available directly from Meta
It requires expensive infrastructure Retired with on-prem Cloud API is hosted by Meta at no hosting cost

Two of these matter enough to plan around. The 1 October change turns your highest-volume message type from free into billable, and it is covered in detail below. The token-billed AI agent is already live, already charging, and is discussed in its own section, because the unit change hides how large the number is.

The rest matter mostly because they distort budgets. A business that has modelled its costs on conversation pricing has usually underestimated a chatty support workload and overestimated a low-touch notification workload, since the old model bundled an unlimited number of messages into one 24-hour price and the new one does not.

Access is granted, not purchased

The most useful mental correction: you are not buying software with a licence key. You are being granted metered access with a behaviour-linked credit limit, and every part of that grant is adjustable by Meta after the fact.

Four separate controls sit on your account, and each moves independently:

Messaging limit. How many unique people you can start a conversation with in a rolling 24 hours. New numbers begin at 250. It rises through 1,000, 10,000 and 100,000 to unlimited, based on volume and quality, and it falls when quality drops.

Quality rating. A rolling green/yellow/red signal per number, driven by how recipients react — blocks, reports, and whether they reply at all. Red precedes a limit reduction.

Template status. Each pre-approved message you send outside the window is individually approved, individually categorised, and can be paused mid-campaign if recipients react badly to that specific template.

Account status. Business verification, display-name approval and the official green tick each have their own review, and each can be revoked.

None of these are things a provider can give you or protect you from. A good provider surfaces them clearly; a bad one hides them until something breaks. When you evaluate platforms, ask what you can see, not what they promise.

This is also why the phrase “we guarantee no bans” should end a sales conversation. No reseller controls any of the four dials above.

What you need before you can send anything

Five things, and only the third takes real time.

A Meta Business Account — the container that owns your assets. Most businesses already have one from running ads or a Facebook Page.

A phone number you can dedicate — able to receive one verification code, and then not used in the WhatsApp app or WhatsApp Business app again. It can be a landline. It should not be the number your customers already message if you cannot afford that inbox to go quiet during migration.

Business verification — Meta checks that your business is real, using registration documents, a website and a matching address. This is the slow gate: commonly a few days, sometimes two weeks or more if a document mismatches. Start it first, not last. You can send in a limited way before it completes, but your messaging limit stays at the bottom tier until it does.

A display name that survives review — the name recipients see. Meta rejects names that are generic, misleading, or unrelated to your verified business. “Sales Team” fails. Your actual brand passes.

A provider or a direct Cloud API setup — covered next.

Nothing here costs money at Meta’s end. Verification is free, the Cloud API is free to access, and the green tick is free to apply for. What you pay is per message, plus whatever your provider charges.

Cloud API, on-premises, or a BSP

Three ways to be on this API, and the choice is less consequential than it used to be.

Cloud API is Meta hosting everything. No servers, no upgrades, current features first. This is now the default, and a small business has no serious reason to choose otherwise.

On-premises API was Meta shipping you a container to run yourself. It has been deprecated and is being wound down; new businesses cannot start here, and the residual reasons — data residency, a compliance regime that forbids message content leaving your infrastructure — are narrow and shrinking. If a provider is still selling it to you as an advantage in 2026, ask why.

A Business Solution Provider (BSP) sits between you and the Cloud API. You pay them; they handle onboarding, sometimes billing, and they give you the interface Meta does not — inbox, agents, bot builder, reporting, integrations. Almost everyone is on this path, whether or not the word BSP appears on the invoice.

Worth separating one thing here, because the same search term collects two very different readers: this guide is for businesses that want to use the API. If you want to sell it — become a BSP or a Tech Provider and earn on other people’s message volume — that is a different business with different economics, and it is covered in the WhatsApp Business API reseller guide.

The practical difference is not technical. Going direct to the Cloud API costs you Meta’s rate and nothing else, but you need a developer to build or wire up everything a person actually uses. Going through a provider costs a platform fee, often a markup on Meta’s rate, and buys you the layer. For most businesses the platform fee is smaller than a fortnight of developer time, so this is not a close call — but you should know which part of your bill is Meta’s and which is theirs. A provider that will not itemise that split is telling you something.

If your problem is the layer, not the plumbing, Lion CRM gives one number a shared pipeline, labels, notes and saved replies through a Chrome extension — running this afternoon, no verification queue, no per-message fee. See the platform.

What setup really looks like

Timelines quoted in marketing material — “live in 10 minutes” — measure the last step only. Here is the honest sequence.

Days 1–2: paperwork. Meta Business Account created or claimed, business verification submitted with registration documents that match your website and address exactly. Mismatches are the main cause of delay. A number chosen and freed from any existing WhatsApp app.

Days 2–10: waiting, with work you can do. Verification is in review. Meanwhile, choose and connect your provider, register the number, get the display name approved, and draft templates. Templates can be submitted and approved during this window, which is why drafting them early is worth doing.

Days 5–12: templates and the first real send. Template approval is usually hours, occasionally a day. Category matters more than wording: a message Meta reads as promotional is billed as marketing even if you meant it as a notification.

Weeks 2–4: warm-up. You start at 250 unique conversations per 24 hours. Volume and consistently good quality move you up. There is no way to buy your way past this, and trying to force volume on a cold number is the reliable way to end up at red quality with a reduced limit.

Budget three to four weeks before the API is doing real work at scale. Businesses that plan for ten minutes and get three weeks are the ones who describe onboarding as painful.

Three meters, and a fourth in October

This is the section that most changes what a WhatsApp budget should look like. Meta currently runs three different meters against one API, and they do not share a unit.

Meter one — template messages, per message, by category. Marketing, utility and authentication each have their own country rate. You pay when the message is delivered.

Meter two — service messages, currently free. Free-form replies inside the 24-hour window after a customer writes to you. Free since 1 November 2024. Billable from 1 October 2026 at the same rate as utility and authentication in each country.

Meter three — Meta Business Agent, per token. Live since 1 August 2026. Replies generated by Meta’s own AI agent are billed at $2.00 per million tokens, bundling AI processing and delivery into one charge. A typical delivered message consumes roughly 20,000 to 25,000 tokens.

The fourth arrives on 1 October when meter two starts charging. It is not a new mechanism — it is meter one’s rate applied to a message type that has been free for two years.

Two things follow. First, “cost per message” is now a meaningless number unless you say which meter. Second, the cheapest message and the most expensive message on this platform differ by more than an order of magnitude, and the expensive one is the one that looks most like the future.

What it costs in India right now

Meta’s India rates, per delivered message, as of August 2026:

Category Meta’s rate (INR) When you pay it
Marketing ₹0.8631 Any promotional template, in or out of the window
Utility ₹0.115 Order, delivery, payment, appointment templates
Authentication ₹0.115 OTP and verification templates
Service Free until 1 Oct 2026 Free-form replies inside the 24-hour window
Meta AI agent reply ~₹3.50–₹4.40 Any reply generated by Meta Business Agent

Meta publishes these as rate cards by currency, and the current generation took effect 1 July 2026. Check the official pricing documentation before you commit a budget — these numbers move.

Three layers sit on top of Meta’s rate before it reaches your card:

Your provider’s platform fee. In India this commonly runs ₹999 to ₹17,000 a month depending on tier, agent seats and features.

Your provider’s markup on Meta’s rate. Frequently ten to thirty percent. Some providers pass Meta’s rate through untouched and earn on the platform fee instead. Ask which model you are on, in writing.

GST at 18%, on both.

A worked line: 10,000 utility messages a month is ₹1,150 at Meta’s rate, ₹1,265 to ₹1,495 after a typical markup, plus a platform fee of say ₹2,999, plus 18% GST on the total — roughly ₹5,000 to ₹5,300 all in. The Meta component is under a quarter of the bill. That ratio flips as volume grows, which is why per-message markup matters far more at scale than the headline plan price. Our full India rate-card breakdown works through the three layers with more examples.

Meta’s own AI is the priciest message on the platform

Put the numbers next to each other, because nobody selling this platform will.

A utility template in India costs ₹0.115. A marketing template costs ₹0.8631. A reply generated by Meta Business Agent, at $2.00 per million tokens and roughly 20,000 to 25,000 tokens per delivered message, costs about four to five US cents — call it ₹3.50 to ₹4.40 at August 2026 rates.

That is roughly thirty to forty times a utility message, and about four to five times a marketing message. For a single reply, in a conversation the customer started, which would have been free under meter two.

The strategic consequence is not “avoid AI”. It is that AI replies must now be justified per message rather than per feature. A bot that answers a genuine question and closes a support ticket is easily worth ₹4. A bot that opens with “Hi! How can I help you today?” and then asks a clarifying question before doing anything useful has burned ₹8 before it has said anything, and it does that on every conversation.

Three practical rules fall out of this:

Route the cheap questions away from the model. Order status, hours, address and price-list requests are keyword or menu jobs. They should never reach a language model, on any platform. This is the same argument for a rules-based flow that has always existed, but the price gap now makes it arithmetic rather than taste.

Make the first AI reply carry the whole answer. Every clarifying turn is another full charge. A bot that resolves in two messages costs half what an identical bot resolving in four costs, at identical quality.

Know whose model you are paying for. Meta’s token billing applies to Meta Business Agent. A bot built on your provider’s platform, or your own model behind the API, is billed by that provider or that model vendor, not by Meta’s token meter — though the underlying message still goes through meters one or two. Some of those arrangements are considerably cheaper; some are not. Ask for the per-resolved-conversation number, not the per-message one. We work through when a bot earns its place in the WhatsApp AI chatbot guide.

1 October 2026, and what to do before it

On 1 October 2026, service messages stop being free. Utility templates sent inside the 24-hour window lose their in-window free status at the same time. Meta has said it will publish exact rates by 1 September 2026, and service messages will be priced at the same per-message rate as utility and authentication in each country — which in India would put them at ₹0.115 unless the rate card changes.

Note what this does and does not touch. It applies to the WhatsApp Business Platform only. The free WhatsApp app and the free WhatsApp Business app are unaffected, which quietly makes the non-API routes cheaper in relative terms for the first time in years.

The businesses most exposed are the ones whose workload is mostly replies: support desks, shared inboxes, anything with a bot answering inbound questions. That traffic has been free since November 2024 and has never appeared on a cost line.

Six weeks is enough to prepare properly:

Measure messages per resolution now. Pull your last 30 days and calculate the average number of outbound messages per conversation that reached a conclusion. That single number is your October multiplier. Most teams have never looked at it, and most are surprised.

Find the padding. “Sure!”, “One moment”, “Let me check for you”, the standalone “Anything else?” — every one of those is a free message today and ₹0.115 in October. On 20,000 conversations a month, cutting two padding messages per conversation is ₹4,600 a month.

Compress bot flows. A five-step menu tree that ends in an answer costs five messages. A flow that asks one question and answers costs two. The design change is small; the bill difference is 60%.

Re-examine which conversations need the API at all. If a workload is a person replying to inbound messages during business hours, on one number, that is what a browser extension over WhatsApp Web does for a flat monthly fee and no per-message charge, on any volume. Post-October, that gap widens.

Get a written answer from your provider on billing. Ask specifically: will service messages be passed through at Meta’s rate, or marked up like templates? Some providers will absorb it in the platform fee. Ask before September, when everyone is asking.

Templates: the part everyone underestimates

Outside the 24-hour window, you cannot type freely to a customer. You can only send a template that Meta approved in advance, and templates carry more operational weight than any feature list suggests.

Category decides price, and Meta decides category. You propose marketing, utility or authentication; Meta’s review decides. A “your order has shipped” message with a discount code attached is re-categorised as marketing, and the difference is ₹0.115 versus ₹0.8631 — a 7.5x swing on your highest-volume template. Keep promotional content out of utility templates, not to game the system but because bundling them is what triggers the re-categorisation.

Variables are constrained. You define placeholders; you cannot restructure the message at send time. Any change to fixed text means a new submission and a new review.

Approval is fast until it is not. Usually within hours. Rejections cluster around unclear variables, promotional content in a utility category, and anything reading like it came from a purchased list.

Templates are paused individually. A single template that draws blocks and reports can be disabled while the rest of your account continues normally. This is a mercy, not a punishment — it is how Meta avoids penalising the whole number for one bad campaign.

The practical implication is that template design is a recurring job, not a setup task. Teams that treat it as setup end up sending a badly categorised template ten thousand times a month and paying seven times too much for a year.

The 24-hour window is a billing boundary

When a customer messages you, a 24-hour window opens. Inside it you can reply in free-form text. Outside it you can only send approved templates.

Until 1 October 2026, that window is also a price boundary: inside is free, outside is billed. After 1 October it stops being a price boundary and becomes purely a format boundary — inside you can write what you like, outside you must use an approved template — with both sides billed.

This is a bigger design change than it sounds. Almost every piece of WhatsApp advice written since November 2024 optimises for getting inside the window and staying there, because inside was free. That advice expires. What remains true is the format constraint: free-form conversation only works inside 24 hours, so response time still governs whether you can have a normal conversation at all.

One thing that does not change: a customer’s message reopens the window. If your reply is slow enough that they message again to chase you, you have not saved anything — you have added a message to your own bill and a poor experience to theirs.

Measuring messages per resolution is easier when you can see the conversation, which is the job Lion CRM does on top of your own WhatsApp Business Account — pipeline, assignment and history on the extension route and the official-API route alike. See the platform.

Messaging limits and quality rating

Two controls, frequently confused.

The messaging limit caps how many unique customers you can initiate conversations with in a rolling 24 hours. Tiers are 250, 1,000, 10,000, 100,000, then unlimited, and Meta documents the escalation rules in its messaging limits reference. You move up by sending consistently at good quality. Replies inside an open window do not count against it — the cap is on conversations you start.

The quality rating is a rolling per-number signal shown as green, yellow or red, derived from recipient behaviour. Blocks and reports pull it down hardest. Being ignored also hurts: a template nobody responds to reads as unwanted.

They interact. Red quality can reduce a limit that took months to earn, and recovery takes weeks of clean sending. This is why a business that hits a hard cap mid-campaign should slow down rather than push — the cap is protecting a rating whose loss costs more than the delayed campaign.

For most small businesses none of this is a live constraint. A team having conversations customers started never approaches 250 initiations a day. It becomes real the first time you run a broadcast to your full list, which is exactly when people discover it.

What actually gets you restricted

The persistent myth is that Meta bans numbers for using unofficial tools. It does not, in the sense people mean. Enforcement tracks sending behaviour, and the signals are the same whichever route you send on.

What actually moves the needle: recipients blocking you, recipients reporting you, high volumes of outbound to people who never contacted you, identical message bodies at machine speed, and near-zero reply rates. Those five together describe cold blasting, and cold blasting is what gets restricted.

What does not, by itself: using a browser extension, using the free Business app, or sending high volume to people who genuinely want to hear from you.

The official API does not exempt you. It makes volume legitimate — you can send 100,000 initiations a day at the top tier — but every one of those messages is measured for quality, and the enforcement response is faster and better instrumented than on any other route, because Meta can see everything. Businesses that move to the API expecting a shield are the ones most surprised by their first quality downgrade. The mechanics, route by route, are in our guide to sending bulk WhatsApp messages without getting banned.

Do you actually need the API?

Here is the test nobody selling the API will give you. If none of these describes you, you do not need it, and adopting it will add cost and setup time for capability you will not use.

You need the API if:

  • Messages must be triggered by another system — your store, your CRM, a payment gateway, a booking tool.
  • Messages must fire when nobody is at a computer: 3am order confirmations, appointment reminders on a Sunday.
  • Several agents must share one number with real assignment, so two people never answer the same customer.
  • You send template messages at volume to people outside the 24-hour window.
  • You need a chatbot that runs server-side and does not depend on a browser tab.
  • You need the green tick, which requires the API plus a separate brand review.

You do not need the API if:

  • One or two people handle enquiries during business hours.
  • Everything you send is a reply to something a customer sent first.
  • Your volume is in the low hundreds of conversations a month.
  • What you actually want is to stop losing track of who asked what — that is a CRM problem, not an API problem.

That last one is the most common. A large share of businesses that go through API onboarding wanted visibility — a shared view of open enquiries, who owns each, what was promised — and bought a messaging channel instead. A Chrome extension over WhatsApp Web gives one number a shared pipeline, labels, notes and saved replies with no per-message fee and no template approval, and it can be running this afternoon. We compare the two shapes directly in WhatsApp CRM vs WhatsApp Business API.

The honest sequence for most small teams is: fix visibility first on the route you are already on, then move to the API when a trigger appears that a human at a keyboard genuinely cannot serve.

What the API does not give you

Worth stating plainly, because the omissions cause the most disappointment.

No inbox. Nowhere to read or reply as a person. Your provider supplies this.

No contact management. No pipeline, no stages, no notes, no history beyond the message log. Provider or CRM.

No reporting. Delivery and read receipts arrive as webhook events. Anything resembling a dashboard is built on top.

No bot builder. The API accepts and delivers messages. Logic is yours or your provider’s.

No content protection. Approval covers structure, not judgement. A template can be perfectly approved and still be the wrong thing to send.

No exemption from the rules. Covered above, and worth repeating because it is the belief that most reliably costs money.

This is why “should I get the API or a CRM” is a malformed question. The API is plumbing. A CRM is a product. Nearly everyone with the API also has something on top of it — the only question is whether they chose that layer deliberately or accepted whatever their provider bundled. Our WhatsApp CRM software guide covers what that layer should contain.

Choosing a provider: eight questions

Ask these in writing, before you sign. The answers separate providers far better than any feature grid.

  1. Do you mark up Meta’s per-message rate? By how much, per category? The single biggest cost variable at volume, and the one most often left out of a quote.
  2. What will you charge for service messages after 1 October 2026? Anyone without an answer in August has not planned.
  3. Is the WhatsApp Business Account in my Meta Business Manager, or yours? If it is theirs, your number, templates and history are hostage to the relationship. Insist on yours.
  4. Can I see my quality rating and messaging limit without asking you? Hidden means you find out about a downgrade from a failed campaign.
  5. What happens to my conversation history if I leave? Meta’s migration moves the number, templates, quality rating and tier in a few days. It does not move conversation history, pipeline stages, bot flows or reporting — those have no defined format and generally do not travel.
  6. Is the platform fee per agent seat or per account? Seat-based pricing is fine until you add a fifth agent.
  7. What is included in support, and what is the actual response time? Get a number and hours, not “dedicated support”.
  8. Which of these features are yours and which are Meta’s? Some providers describe Meta’s platform capabilities as their own. A quick test: ask what happens to the feature if you switch providers.

If a provider will not itemise the Meta rate separately from their own charges, treat that as the answer to question one.

Three businesses, three real bills

India, August 2026 rates, before the October change. Provider fees are mid-market; markup assumed at 15%; GST 18% on the total.

A 4-person clinic. 600 inbound enquiries a month, all replies inside the window, 400 appointment-reminder utility templates. Meta: ₹46 (templates) + ₹0 (service). Platform fee ₹2,999. Total with GST, about ₹3,600. After October, the reply traffic — roughly 2,400 messages — adds about ₹276 before markup and GST, so call it ₹3,975. The honest observation here: this clinic’s entire Meta bill is under 2% of what it pays, and its real question is whether it needs the API at all, or a shared inbox on the number it already uses.

A 12-person D2C brand. 8,000 marketing templates, 12,000 utility (order and delivery), 30,000 service replies. Meta: ₹6,905 + ₹1,380 = ₹8,285, plus 15% markup = ₹9,528. Platform fee ₹8,999. With GST, about ₹21,850. After October, 30,000 service messages at ₹0.115 plus markup adds roughly ₹3,970 with GST — a 19% increase, arriving in one billing cycle. If Meta’s AI agent were answering a third of those replies instead of a rules-based flow, the same volume would cost around ₹39,000 on the token meter alone.

A 40-person services firm. 60,000 service replies, 5,000 utility, no marketing. Meta today: ₹575. Platform fee ₹17,000, seat-based. With GST, about ₹21,500. After October, 60,000 service messages add roughly ₹7,935 with GST — a 37% increase on a business that currently thinks of WhatsApp as a fixed cost. This is the profile that should be measuring messages-per-resolution this month.

The pattern: the more your workload looks like conversation rather than campaigns, the harder October lands, and the less any of it showed up on your bill before.

A 30-day plan

Week 1 — decide whether you need it. Run the test above honestly. If nothing in the “you need it” list applies, stop here and fix visibility on your current route instead; you will save three weeks and a recurring bill. If something does apply, start business verification on day one, because it is the only step you cannot compress.

Week 2 — set up while you wait. Choose a provider with the eight questions answered in writing. Register the number, get the display name approved, draft your first three templates — an order or booking confirmation, a reminder, and one genuinely useful marketing message. Keep promotional language out of the first two.

Week 3 — send small and measure. Start well under the 250 cap. Watch quality daily. Measure two numbers from the first day: messages per resolved conversation, and reply rate per template. The first is your October exposure; the second predicts your quality rating before Meta tells you.

Week 4 — build the two automations that pay. An instant first reply, and a two-touch follow-up at day two and day five for enquiries that go quiet. Those two return more than everything else combined. Add a bot only after you can name the eight questions it will answer, and route those eight to a menu rather than a model unless a genuine language problem exists.

Then, before 1 October: revisit messages per resolution, cut the padding, compress the flows, and get your provider’s service-message pricing in writing. A business that does those four things enters October with a bill it has already modelled. Most will not, and will find out in the first week of November.

Want the layer that sits on top? Lion CRM runs on your own WhatsApp Business Account, with a Chrome extension for teams still working in WhatsApp Web and a full official-API platform for when the triggers move to the server. See the platform.

Frequently asked questions

What is the WhatsApp Business API?
It is a programmatic channel into WhatsApp with no interface of its own. It accepts messages from your software and delivers them, and it streams delivery events back. Everything a person uses — inbox, agent assignment, notes, reporting, bot builder — is built on top by a provider. It runs on a server rather than a phone, which is why it can fire when nobody is at a computer, and it attaches to a single number that can no longer be used in the WhatsApp or WhatsApp Business apps.

How much does the WhatsApp Business API cost in India?
Meta charges per delivered message: ₹0.8631 for marketing, ₹0.115 for utility and ₹0.115 for authentication as of August 2026. Free-form replies inside the 24-hour window are free until 1 October 2026. On top of Meta’s rate you pay a provider platform fee, commonly ₹999 to ₹17,000 a month, often a markup of ten to thirty percent on the per-message rate, and 18% GST on both. At low volume the platform fee dominates the bill; at high volume the markup does.

Is the WhatsApp Business API free?
Access is free — there is no licence fee, business verification costs nothing, and the Cloud API is hosted by Meta at no hosting charge. Messages are not free. You pay per delivered template message by category, and from 1 October 2026 you also pay for free-form replies inside the 24-hour window. Almost everyone also pays a provider for the interface, because the API has none.

What changes on 1 October 2026?
Service messages — the free-form replies you send inside the 24-hour window after a customer writes to you — become billable, at the same per-message rate as utility and authentication in each country. Utility templates sent inside that window lose their free status at the same time. Meta has said it will publish exact rates by 1 September 2026. Only the WhatsApp Business Platform is affected; the free WhatsApp app and WhatsApp Business app are not.

How is Meta’s AI agent billed on WhatsApp?
Since 1 August 2026, replies generated by Meta Business Agent are billed by token at $2.00 per million tokens, bundling the AI processing and message delivery into one charge. A delivered message typically consumes 20,000 to 25,000 tokens, which works out to roughly four to five US cents — about ₹3.50 to ₹4.40. That is roughly thirty to forty times an Indian utility template, so AI replies now need justifying per message rather than per feature.

Do I need the WhatsApp Business API for my business?
Only if you need messages triggered by another system, messages that fire when nobody is at a computer, several agents sharing one number with real assignment, template messaging at volume outside the 24-hour window, a server-side chatbot, or the green tick. If one or two people answer enquiries during business hours on one number, you do not need it — and if your actual problem is losing track of who asked what, that is a CRM problem that a browser extension solves faster and cheaper.

What is the difference between the WhatsApp Business app and the API?
The Business app is a free phone app with a catalogue, quick replies, labels, a greeting message and an away message, used by one person on one device. The API is a server-side channel with no interface, supporting many agents on one number, chatbots, messages triggered by other systems, and template messaging at volume. The app is free with no per-message charge; the API charges per message and normally comes with a provider fee.

How long does WhatsApp Business API approval take?
Plan for three to four weeks to be doing real work at scale. Business verification is the slow gate — commonly a few days, sometimes two weeks or more if your registration documents do not match your website and address exactly. Number registration and display-name approval are quick, template approval is usually hours, and then a warm-up period moves you off the starting limit of 250 unique conversations per 24 hours.

Can my number get banned on the official API?
Yes. The API makes volume legitimate, not exempt. Meta enforces on sending behaviour — blocks, spam reports, cold outbound to people who never contacted you, identical bodies at speed and near-zero reply rates — and it can see all of it more precisely on the API than anywhere else. Enforcement usually arrives as a quality-rating downgrade and a reduced messaging limit before anything harsher.

Can I move my number to a different provider?
Yes. Meta’s migration process preserves the number, approved templates, quality rating, display name and messaging tier, typically in three to seven days with minutes of real downtime. What does not travel is the layer on top: conversation history, pipeline stages, chatbot flows, routing rules and reporting history have no Meta-defined format. Ask a prospective provider what leaves with you before you sign, not after.

Become a Reseller