WhatsApp Business API free: what actually costs nothing

Meta's documentation is unambiguous on the unit: "Effective July 1, 2025, Meta charges on a per-message basis." Before that date the platform billed per 24-hour conversation, so a business could send six templates inside one window and pay once. That arithmetic is dead, and any budget, spreadsheet or agency proposal still built on conversation counts is measuring the wrong thing. If your model still speaks in conversations, translate it first - the cost-per-conversation metric is still useful for reporting, but it is no longer the billing unit.

Everything you send falls into one of four buckets, and the bucket sets the price:

• Marketing templates - promotions, offers, cart recovery, re-engagement. The most expensive category in every market, with no in-window exemption.
• Utility templates - order confirmations, delivery updates, appointment reminders, payment receipts. A fraction of the marketing rate. Free inside an open customer service window until 30 September 2026; billed there too since 1 October 2026.
• Authentication templates - one-time passcodes and login verification. Priced separately again, and in some markets a distinct international authentication rate applies.
• Service messages - free-form replies sent inside an open 24-hour window. Free from November 2024 to 30 September 2026, and billed per message since 1 October 2026 at the market's utility and authentication rate.

The second variable is geography, and it is the one people get wrong most often. Rates vary by the template category and by the recipient WhatsApp phone number's country calling code. Where your company is incorporated and where your servers sit are irrelevant. A German retailer messaging customers in India pays Indian rates on those messages - so a single blended "cost per message" figure is meaningless for anyone with an international audience.

There is a third variable that reduces cost rather than raising it. Meta applies volume tiers to utility and authentication messages: send more in a calendar month and the per-message rate for those two categories falls. Volume is aggregated across every WhatsApp Business Account in your business portfolio and the tiers reset monthly. Marketing is not tiered. The mechanics, along with the tier thresholds that govern how many unique people you may message at all, sit on the WhatsApp limits reference alongside the messaging tier ladder.

We are deliberately not printing a rate table on this page. Per-country rates change on announced dates and per category, and a table copied into a blog post in March is a liability by September - which is exactly how the rest of this search result became unreliable. Instead, put your own volumes into the WhatsApp API cost calculator, which carries editable indicative rates stamped with the date they were last checked, and cross-check against Meta's own rate card before committing a budget. For just the free-versus-billable split, the calculator's free-tier breakdown answers that in isolation.

One last cost that is not a message at all: from 1 August 2026 Meta charges per token for Meta Business Agent replies, at one global rate of $2.00 per million tokens. If you are evaluating Meta's own AI layer, that is a second meter running beside the message meter.

This is the part of the answer that dates every other page on this query. Meta's pricing update notice states: "Effective October 1, 2026, Meta will charge for service messages, which have not been charged since November 2024." A second line in the same notice closes the other free door: "Effective October 1, 2026, Meta will charge on a per-message basis for utility messages sent in response to users within an open 24-hour customer service window."

Together they removed the largest genuinely free thing about the WhatsApp Business API. For nearly two years a business that only answered inbound questions paid Meta almost nothing: the customer wrote first, the 24-hour customer service window opened, and every reply inside it - typed by a person or generated by software - was free. Since 1 October 2026 those replies are billed individually, at the same per-market rate the country already pays for utility and authentication messages. The rates are on your own rate card in WhatsApp Manager.

Three things are worth separating, because they get conflated constantly:

• This is not a rate rise. The published price of a marketing, utility or authentication template is not what changed. A category that cost nothing simply entered the meter.
• Inbound is still free. You are not charged when a customer messages you. You are charged for what you send back. A high-traffic support number does not become expensive because it is busy; it becomes expensive because it is chatty.
• Free entry points survive. Meta's notice is explicit that messages sent within the 72-hour free entry point window - opened from Click to WhatsApp Ads or Facebook call-to-action buttons - remain free for message delivery. That is now the only large free window left.

For budgeting, the consequence is blunt: any WhatsApp budget built before October 2026 is sizing the cheaper of two worlds. A few hundred conversations a month will barely notice. Tens of thousands should re-forecast now rather than discover the new line in a November invoice, and the variable that now matters is messages per resolution - a four-message resolution now costs twice a two-message one, which was never true before.

The full arithmetic, the industries most exposed, and the decisions worth making now are worked through in our guide to the October 2026 service message billing change. To put your own numbers against it, the before-and-after view of the cost calculator shows both eras side by side. And if you have never seen the window close on you in production, the error it produces is 131047 Re-engagement message.

Meta's rate is the floor. If you send through a Business Solution Provider - which includes us - there is a second bill, and the shapes it comes in matter more than the names attached to them. There are three, and most vendors blend them:

1. Per-message markup. A fixed amount added to Meta's rate on every billable message, sometimes hidden inside a rounded-up "our rate" table that never shows Meta's underlying figure. It scales linearly with volume: cheapest for low senders, most expensive for high ones.
2. Monthly platform fee. A subscription for the software - builder, inbox, analytics, team seats - with Meta's message charges passed through at cost. Predictable, and it gets cheaper per message as you grow.
3. Both, tiered. A monthly fee including an allowance of conversations or messages, with overage priced per unit above it. Value depends entirely on how close your real volume sits to the allowance boundary.

None of these is dishonest in itself. What makes a quote hard to evaluate is the two lines being merged, so you cannot tell whether a rate rose because Meta raised it or because your vendor did. Four questions separate them cleanly, and they work on any vendor including this one:

• Is Meta's rate shown separately on the invoice? If not, you cannot audit either line or see what the October 2026 change added to your bill.
• How are service messages billed since 1 October 2026? A vendor with no published answer has not thought about your bill.
• What is the unit of the allowance? "Conversations" and "messages" have not been the same thing since July 2025. Ask which one the plan counts.
• What do you keep if you leave? Your number, your WhatsApp Business Account, your approved templates and your history should be portable - that is a Meta-level concept, not a vendor favour.

For completeness on our own side: Conferbot charges a monthly platform fee and does not add a per-message markup on top of Meta's rate - the plans are listed on the pricing page, and the free tier is described precisely further down. If you want to model a markup - ours, a competitor's, or a hypothetical - the cost calculator has a per-message markup field that applies to billable messages only. If you would rather integrate against a plain HTTP surface than a vendor console, our chat API and API integration features are the relevant pages, and agencies reselling under their own brand should look at white-label.

Yes. This is the section most pages on this query omit, because it argues against the page's own product. Meta's Cloud API is available directly: no approval queue you can only join through a partner, no reseller-only pricing, no capability a Business Solution Provider holds exclusively. If you have an engineer and a genuinely simple use case, going direct is a legitimate answer and frequently the right one.

What "direct" actually requires, in the order you will hit it:

1. A Meta business portfolio and business verification. The portfolio is the container that owns your WhatsApp Business Account. Verification is a document-review process, not a form, and it gates your ability to move past the starting messaging tier - the ladder from 250 unique recipients per rolling 24 hours upward is documented on the messaging tiers reference.
2. A Meta app and a WhatsApp Business Account. Straightforward, and where Meta provisions a test business phone number for you.
3. A phone number you control that is not already on WhatsApp. This is where most first attempts stall. A number attached to the consumer WhatsApp app or the WhatsApp Business app has to be deleted from it first, and a number that cannot receive the verification code is a dead end. The specific failures are catalogued in why WhatsApp number registration fails, with the constraints on phone numbers and registration.
4. An HTTPS webhook endpoint with a valid certificate. Meta will not deliver to plain HTTP, to a self-signed certificate, or to an endpoint that fails the verification handshake. You own the signature check, the 200-within-the-timeout contract, and de-duplication of retried deliveries - retry behaviour is on the webhooks reference, the concept itself in the webhook glossary entry, and the debugging path when nothing arrives in the webhook not firing guide.
5. Template submission and category management. Every business-initiated message is a template, every template is reviewed, and the category Meta assigns determines the price. A template you wrote as utility can be reclassified as marketing and quietly multiply its cost. Run drafts through the template checker and read why WhatsApp templates get rejected before you submit a batch. Structural limits - variable counts, button counts, body length - are on the template limits reference.
6. Your own error handling. This is the part that is invisible in a proof of concept and dominant in production. You need code paths for 131047 when the window has closed, 131026 when the recipient cannot receive, 130429 when you exceed throughput, 131048 for quality-based throttling, 131049 for per-user marketing caps, and 190 for expired tokens. The full catalogue is at WhatsApp Cloud API error codes, with the practical walkthrough in the error codes guide.
7. Quality monitoring. Your number carries a quality rating that can throttle or restrict it. Recovering from a restriction is a process, not a retry - see quality rating and phone number status and what to do when a WhatsApp Business account is restricted.

None of that is exotic, and all of it is real work: two or three days to a working demo, then a recurring maintenance obligation that does not go away. The step-by-step version with the Graph API calls is in the setup guide; the vocabulary is in the API endpoint and WhatsApp Business API glossary entries; and every documented rate limit, cap and tier is on the WhatsApp limits page.

Having argued for going direct, here is the honest boundary. Self-serving the Cloud API is fine when the scope is narrow and stays narrow: one number, a handful of templates, a developer who owns it, and no non-technical person who needs to change a message on a Friday afternoon. Under those conditions a platform is overhead.

It stops being fine at a set of recognisable points:

• Someone non-technical needs to change the conversation. The moment a marketer has to file a ticket to edit a reply, your flow is a bottleneck rather than a channel. A visual builder and a library of ready flows is what that person needs.
• You add a second channel. Once the same conversation has to work on the web widget or Telegram too, you are building routing and session state yourself - what omnichannel handling across the supported channels absorbs. Trade-offs in choosing a channel strategy.
• A human has to take over mid-conversation. Handoff, presence and history are deceptively hard to build, and are what live chat handles.
• The 24-hour window is a cost centre. Since October 2026, messages per resolution is a budget line, and measuring it needs instrumentation you would otherwise build yourself.

Conferbot's free tier, stated precisely. It is a free tier, not a trial: it does not expire, there is no countdown, and it does not ask for a credit card. It includes 600 conversations a month, one bot and two seats. Above that you move to a paid plan - $19, $39 or $59 a month - or you wait for the month to roll over. Meta's per-message charges are separate and are billed by Meta, not by us; no platform can waive them, and any page implying otherwise is selling you something. The full breakdown is on the free plan page and the pricing page, and what the WhatsApp channel itself does is on the WhatsApp chatbot page. Industry-shaped versions exist too, for ecommerce, healthcare, logistics and insurance, and the wider connector list is at integrations.

Three different things get called "free WhatsApp API" and they are not comparable: Meta's own free message types, a vendor trial with an expiry date, and a permanent software tier with a volume ceiling. The table describes the models rather than quoting competitor numbers - a price scraped into a table today is wrong within a quarter, and we are not a neutral party about anyone else's figures. Check any vendor's own pricing page on the day you evaluate them.

The row that surprises people is the first one. Meta's free elements are the largest of the three by a wide margin and are available to everybody, including customers of every vendor on this search result. No platform grants them and none can take them away. What a platform changes is how much engineering you spend to use them, and - through response latency and conversation design - how often your traffic lands in the free categories at all. That second effect beats any discount: a bot answering in two seconds keeps windows open that a team replying next morning has already let expire.

Compare vendors on the two lines separately - software cost, and markup per message. Put both into the cost calculator at your real volumes and the ranking often inverts between low and high senders. If you are still deciding whether WhatsApp is the right channel at all, start with the free WhatsApp chatbot guide or conversational commerce.