WhatsApp Chatbot Cost: What Drives the Price of an AI Agent
By Haseeb Asad, founder of Codex Labs. Published . Facts checked on 26 September 2026.
A WhatsApp AI chatbot costs a build fee plus monthly running costs. At Codex Labs, full WhatsApp automation starts from $1,499. Where a project lands above that depends on the number of flows, the systems it connects to, voice notes, languages and the knowledge base. Running costs are AI usage, hosting and, on Meta's API, template fees.
Many "WhatsApp chatbot cost" pages give a range so wide it tells you nothing. We would rather not. We have one published price, and below is what actually moves a quote up from it, based on the agents we run in production.
How much does a WhatsApp AI chatbot cost?
Full WhatsApp automation at Codex Labs starts from $1,499, quoted after a short scoping call. That is the build. You also pay running costs every month, and it is worth knowing both before you compare quotes.
| Cost | When you pay it | What sets it |
|---|---|---|
| Build | Once | Flows, integrations, voice, languages, knowledge base, connection type |
| AI model usage | Monthly | Number of messages and how long the conversations run |
| Hosting | Monthly | Self-hosted n8n and database, or n8n Cloud |
| Meta template fees | Per message | Only on the official Cloud API, only for template messages |
| Upkeep | As needed | Menu and price changes, new flows, model or rule changes |
What drives the build price?
These are the drivers we see when we scope a WhatsApp agent.
How many flows the agent runs
Answering questions from a knowledge base is one flow. Taking orders with a confirmation step is another. Bookings, lead qualification, reminders and follow-ups are more. Our restaurant agents run questions, orders with a two-step confirmation and handoff, and the larger of the two is a 94 node n8n workflow. A questions-only agent needs far fewer.
What it connects to
Reading a menu from a spreadsheet is simple. Writing orders into a POS, bookings into a calendar with real availability, or leads into a CRM with deduplication takes longer, and depends on how good the other system's API is.
Voice notes
If your customers send voice notes, the agent needs transcription, a speech rewrite step and text to speech to reply by voice. In our builds that is its own branch of about a dozen nodes, including a formatter agent.
Languages
English only is the baseline. Mixed languages, such as English plus Roman Urdu, or Urdu voice replies, need more prompt work and testing.
The knowledge base
Clean, current menus and FAQs load quickly. Scattered PDFs, images of menus and prices that live in someone's head take time to turn into something the agent can search and quote exactly.
Human handoff rules
A basic "talk to a person" switch is quick. Handoff with staff alerts, confirmation before transfer, automatic return to the agent and staff replies from the phone detected automatically, like in our restaurant case study, takes more.
The WhatsApp connection
Meta's official Cloud API needs a business portfolio, a registered number, and approved templates for anything you send first. An unofficial gateway like Evolution API skips that but carries account risk. We build on both and recommend one per project. We explain both in Evolution API vs WhatsApp Cloud API.
What does a WhatsApp AI agent cost to run each month?
AI model usage. Every customer message calls one or more models: in our builds a small router model, the main agent, and sometimes a formatter. Providers bill per token, so cost scales with message volume and conversation length. We use small, fast models (OpenAI GPT-4.1 mini and Google Gemini 3 Flash, in preview) because they keep per-message cost low. Check the current per-token prices on those pages; they change.
Hosting. Self-hosted n8n Community Edition is free; you pay for the server and the database. If you would rather not run a server, n8n Cloud's Starter plan is 20 EUR a month billed annually (24 EUR monthly) for 2,500 executions (n8n pricing, checked September 26, 2026). Every inbound WhatsApp message uses at least one execution, so size the plan to your message volume.
Meta fees, on the Cloud API only. Since July 1, 2025 Meta charges per delivered template message, by country and category. "All non-template messages are free", and utility templates inside an open customer service window are free too (Meta pricing). So an agent that only replies to customers pays Meta little or nothing. Reminders, follow-ups and lead reach-back use templates, and those are charged.
Voice. Transcription and text to speech are billed by the provider per use. Only relevant if you turn voice on.
We estimate running costs with you before the build, from your real message volume.
What does an n8n automation project cost?
The same logic applies to n8n work beyond WhatsApp: you pay for the build by scope, and the platform is the smaller line. Self-hosted n8n is free apart from the server. n8n Cloud starts at 20 EUR a month, and an execution counts a whole workflow run however many steps it has, which keeps long automations cheap to run. If you are still choosing a platform, our n8n vs Zapier vs Make comparison has the current prices for all three.
We have not published prices for other automation work yet. We quote each one after a scoping call.
Is a WhatsApp AI agent worth the cost?
Work it out with your own numbers rather than ours:
- Time. Messages per day, times the minutes a person spends on each, times what that person's hour costs you.
- Staff. Would you need another person to answer WhatsApp as enquiries grow?
- Speed. In a Harvard Business Review study of 1.25 million online leads, firms that tried to contact a lead within an hour were "nearly seven times as likely to qualify the lead" as firms that waited even an hour longer (HBR, March 2011). An agent replies straight away, at any hour.
Those are the lines an agent works on. It replies at any hour, handles menu questions and orders without anyone typing, and alerts staff only when a chat needs a person.
How do you compare WhatsApp chatbot quotes?
Ask every vendor the same questions:
- Which flows are included, and what counts as a new flow later?
- Which systems does it connect to, and who maintains those connections?
- Is the price one-time, monthly, or both? What are the running costs at my volume?
- Official Cloud API or a WhatsApp Web gateway? Who carries the account risk?
- Who owns the workflows and the data if we part ways?
- What happens when the agent cannot answer?
Our answers: a build price starting from $1,499, running costs estimated up front, and a recommendation between the official Cloud API and a gateway such as Evolution API, since we build on both. See the WhatsApp AI agent service for what is included.
Frequently asked questions
How much does a WhatsApp AI chatbot cost?
At Codex Labs, full WhatsApp automation starts from $1,499 for the build. The final quote depends on the number of flows, the systems it connects to, voice notes, languages and the knowledge base. Running costs for AI usage, hosting and any Meta template fees are separate and scale with volume.
What are the monthly costs of a WhatsApp AI agent?
Three things: AI model usage, which scales with messages; hosting for n8n and the database; and Meta's per-message fees if you use the official Cloud API and send template messages. Replies inside the 24-hour customer service window are free on the Cloud API.
Is it cheaper to build a WhatsApp bot or buy a SaaS chatbot?
A SaaS chatbot is cheaper to start and fine for simple FAQ bots. A custom build costs more up front but has no per-seat fees, connects to your own systems and data, and handles flows like orders with confirmation, voice notes and human handoff that most SaaS tools limit.
Does WhatsApp charge for chatbot messages?
On Meta's Cloud API, since July 1, 2025, you pay per delivered template message by country and category. Non-template messages, which covers bot replies inside the 24-hour window, are free. On a WhatsApp Web gateway like Evolution API Meta charges nothing, but the account carries ban risk.
How much does an n8n automation project cost?
The build is priced by scope, like any software project. The platform is the smaller cost: self-hosted n8n Community Edition is free apart from the server, and n8n Cloud starts at 20 EUR a month billed annually for 2,500 executions. Our WhatsApp automations start from $1,499.
Why do WhatsApp chatbot quotes vary so much?
Because 'chatbot' covers everything from a keyword menu to an agent that takes orders, reads your database, understands voice notes and hands off to staff. Ask each vendor which flows, integrations and running costs a quote includes, and who owns the workflows afterwards.