How much does it cost to automate WhatsApp with AI? Our fee, Meta's messages and your model
Automating WhatsApp with AI has three cost lines: the platform's fee, Meta's message charges on the business's own WhatsApp Business Account (each number gets 1,000 free service messages a month, then per message; templates by category and country), and the AI provider's tokens. A reply's tokens are the instructions and history sent in plus the reply, so shorter prompts and capped replies cost less.
easyCoexistence charges one flat fee per connected number and includes automations, AI replies on your own key, the MCP connector with more than 30 tools and 24/7 monitoring, with no run cap and no markup on Meta's messages or your tokens.
easyCoexistence per connected number per month, every feature included, falling across seven bands as you add numbers.
How Meta charges: by message category and the recipient's country calling code, and only when a message is delivered.
Free service messages Meta gives each business number monthly; since 1 October 2026 it charges service messages from the 1,001st.
Default ceiling on an AI reply's length in easyCoexistence automations, adjustable from 50 to 2,000 tokens.
OpenAI's rough rule of thumb for one token of English text, about 0.75 words.
What easyCoexistence adds to Meta's message charges or to your AI provider's token charges.
What are the three costs of automating WhatsApp with AI?
Three bills from three companies, and only one of them comes from easyCoexistence. Our fee is flat per connected number. Meta charges for the messages your number sends, on your own WhatsApp Business Account. Your AI provider charges for the tokens your model reads and writes, on your own key. The three move independently: a month with more conversations raises Meta's bill and the AI provider's, while ours stays the same as long as the count of numbers does.
What does easyCoexistence charge per number?
A flat monthly fee per connected number, falling across seven bands, and every number on the account pays the rate of the band your total falls in: 6 numbers cost 6 times US$ 7, or US$ 42 a month, and 10 numbers cost US$ 70. Every feature is included: automations with no run cap and no extra charge, AI replies on your own key, the MCP connector with more than 30 tools, webhooks and forwarding, and 24/7 monitoring with alerts. The first 3 days are free; a card is saved before the first number connects, and nothing is charged until the trial ends.
How does Meta charge for the messages an automation sends?
Per delivered message, at a rate set by the message category and the recipient's country calling code, billed to the payment method on your own WhatsApp Business Account through Meta's billing. A fixed reply or an AI reply sent inside the 24-hour customer service window is a service message. Meta gives every number a free tier of 1,000 service messages a month, and since 1 October 2026 Meta charges from the 1,001st, while without a payment method service messages stop once the free tier is used. A Send a template step sends a template, charged by its category: marketing, utility or authentication. easyCoexistence never sees that bill and adds nothing to it. This page quotes none of Meta's rates, because they vary by market and change by date: Meta's rate card is the source, and our page on which messages Meta charges for explains the categories.
How many tokens does one AI reply use?
An AI reply is one call to your model. What goes in: your Instructions (up to 20,000 characters), a block of rules easyCoexistence adds of about 400 characters, the chat's earlier messages (10 by default, up to 20, under "Earlier messages it reads") and the customer's new message. What comes out: the reply, capped by "Longest reply (tokens)", 600 by default and adjustable from 50 to 2,000. Providers price input and output tokens separately, per million tokens, and the count for the same text depends on the model's tokenizer, the language and the content, so treat any rule of thumb as a first estimate.
input tokens ~ instructions + added rules + earlier messages + new message
output tokens <= Longest reply (tokens), 600 by default
cost per reply = input tokens x input price + output tokens x output price
AI cost per month ~ AI replies per month x cost per replyHow do you estimate your monthly AI bill?
Count tokens first, then apply your model's prices. Take a bakery whose Instructions run to 3,000 characters. By OpenAI's rule of thumb of about 4 characters per English token, that is about 750 tokens, plus about 100 for the added rules, about 200 for ten earlier messages of 80 characters, and about 20 for the new message: roughly 1,070 tokens in. A 240-character answer is about 60 tokens out. At 2,000 AI replies a month, that comes to about 2.1 million input tokens and 120,000 output tokens; multiply each by your model's price per million tokens on your provider's pricing page. Then replace the guess with real numbers: open a few live runs in the Runs tab and read the in and out token counts on the AI step.
What keeps an AI bill from running away?
Limits that sit in front of the model call, so a chatty customer or another bot cannot spend your key freely:
- AI replies per chat per hour, at most: 10 by default, from 1 to 60. A customer who writes 30 times in an hour gets at most 10 AI replies, and the rest are skipped.
- Longest reply (tokens) caps what comes out, and Earlier messages it reads caps the history sent in.
- A reply from your phone pauses automations on that chat, 4 hours by default, so no tokens are spent while you handle it.
- Only one automation answers a given message, and echoes, statuses, reactions and our own sends never start a run.
- No model call happens when the 24-hour window has closed or the hourly limit is reached: the run is skipped first.
- Meta's retries of the same message are recognized by message id and do not run twice.
Does easyCoexistence charge for automations or AI?
No. Automations come with the number's fee, with no run cap and no AI credits to buy. The AI key is yours: you add it under AI keys in the dashboard, it is checked with a call that spends no tokens, saved encrypted and never shown again, and the provider bills you directly. We never resell model access or mark up tokens. If the provider refuses the key, for example when an OpenRouter balance runs out, the automations that use it pause with "Paused: your OpenRouter key was refused" and you get one email; they start again once you save a key that works.
Common mistakes
- Estimating from a Test it run. A test sends only the sample message, while a live reply also sends up to 10 earlier messages by default, so live input tokens run higher.
- Topping up OpenRouter credits and waiting. Automations paused by the refused call resume only after you save the key again with Test and save.
- Running AI replies on a WhatsApp Business Account with no payment method. Meta stops delivering service messages once the free tier of 1,000 a month is used, so the AI's replies stop reaching customers.
- Pasting a whole catalog into Instructions. They are sent with every reply, so a full 20,000-character prompt adds about 5,000 input tokens to each answer; keep what customers actually ask about.
Your key stays yours, billed by Anthropic, OpenAI or OpenRouter. US$ 9 down to US$ 0.75 per number per month, falling as you add numbers, with the first 3 days free.
easyCoexistence is a multichannel API and MCP server that connects AI agents such as Claude and ChatGPT to WhatsApp, Instagram, TikTok, Facebook and more. On WhatsApp it uses Coexistence: the business keeps using its number in the WhatsApp Business app while the AI works through Meta's official API.
Frequently asked questions
Do I pay for the messages customers send me?
No. Meta charges only for messages from businesses to users, and easyCoexistence has no per-message charge at all. Receiving costs nothing; the reply is where Meta's charge and your tokens come in.
Which model should I pick to keep AI costs down?
The smallest one whose test replies are good enough for your customers. Providers sell small, fast models at lower token prices than their largest ones, and OpenRouter puts many providers' models behind one key. Compare a few with Test it before you turn the automation on.
Does the 3-day trial cover automations and AI replies?
Yes. The trial includes every feature, automations and the MCP connector too. Your AI provider still bills your tokens and Meta still bills its messages during the trial, since neither goes through easyCoexistence.
Can different numbers use different AI keys?
Within one account, each provider has one saved key, used by all automations that pick that provider. Automations can still mix providers, for example Anthropic on one number and OpenAI on another, and each run shows its own token count.
Keep reading
Ready to get started?
Set up WhatsApp Coexistence in minutes, not months. The app keeps working on the phone.
Start Free TrialVerified on