Evolution API vs WhatsApp Cloud API: An Honest Comparison

What is the difference between Evolution API and the WhatsApp Cloud API?

One detail surprises people: Evolution API is a gateway, not only a WhatsApp Web wrapper. Its README says it "supports both the Baileys-based WhatsApp Web API and the official WhatsApp Cloud API." If you connect it through the Cloud API channel, Meta's rules and prices apply exactly as below.

Can you get banned using Evolution API?

Yes, you can. How likely it is depends mostly on what you do with it, and nobody publishes reliable numbers, including us.

What the sources say, word for word:

• WhatsApp's Terms of Service prohibit uses that "involve sending illegal or impermissible communications such as bulk messaging, auto-messaging, auto-dialing, and the like" or "involve any non-personal use of our Services unless otherwise authorized by us", and say WhatsApp "may take action with respect to your account, including disabling or suspending your account."
• Baileys, the library underneath, says: "Use at your own discretion. Do not spam people with this. We discourage any stalkerware, bulk or automated messaging usage."
• Evolution API's README says the web method "relies on the web version of WhatsApp and may have limitations compared to official APIs." We found no ban warning in its docs.

Our reading: an agent that only replies to customers who wrote first, at human volumes, sits far from bulk messaging and auto-dialing. That is how our restaurant agents run. It is still automated, non-personal use of a WhatsApp account that WhatsApp has not authorized for it, so the risk is never zero. Anyone promising "ban-proof" setups is guessing.

Keep a Cloud API migration path ready if the number matters to your business.

How much does each option cost?

Evolution API. The software is free. You pay for a server to run it (it sits next to n8n on ours) and your time to maintain it. Meta charges nothing on a WhatsApp Web connection.

Two licensing details matter for businesses. The license is Apache 2.0 with additional conditions: keep Evolution's logo and copyright in the frontend, and show a usage notice or you may need a commercial license. And from the 2.4.0 release candidates, every instance must be activated against the Evolution Foundation licensing server. Activation is free with no message limit, but it collects an email, phone number, version and instance ID, and sends usage heartbeats. The current stable release, v2.3.7 (December 5, 2025), does not require it.

WhatsApp Cloud API. From Meta's pricing page: "Effective July 1, 2025, Meta charges on a per-message basis." "You are only charged when a template message is delivered." Then:

• "All non-template messages are free." That covers every free-form reply inside the customer service window.
• "Utility templates delivered within an open customer service window are free."
• "All marketing template messages are charged."
• When a customer arrives through a free entry point, such as a click-to-WhatsApp ad, the window stays open for 72 hours and any message type is free.
• Rates vary by country and category. Meta publishes rate cards by currency, and it scheduled rate card changes for several markets on October 1, 2026.

If you go through a Business Solution Provider rather than directly, they add their own fees. Those vary by provider, so compare quotes.

The practical result: a reply-only support or ordering agent on Cloud API costs little or nothing in Meta fees. Reminders, follow-ups and lead reach-back are template messages, and those are what you pay for.

What are the template and 24-hour window rules?

These apply to the Cloud API. From Meta's Business Messaging Policy:

• "You may only initiate conversations using an approved Message Template."
• "You may reply to a user message without use of a Message Template as long as it's within the 24-hour customer service window, which opens and resets with each user message. Outside the 24-hour customer service window, you may only send messages via approved Message Templates."
• "You may use automation when responding during the 24-hour window, but must also have available prompt, clear, and direct escalation paths", including in-chat transfer to a human agent.

Templates fall into marketing, utility and authentication categories. Meta reviews each one, can reject it, and can recategorize an approved template, for example a utility template it decides is marketing, usually with a day's notice.

You also need opt-in. The policy says you may only contact people who gave you their number and "opt-in permission from the recipient confirming that they wish to receive subsequent messages or calls from you."

Evolution API's WhatsApp Web connection has no template system. That is convenient, and it is also the reason it carries risk: nothing stops you from sending what WhatsApp's terms prohibit.

How does onboarding work for each?

Evolution API: deploy it (Docker is the usual route), create an instance, scan a QR code with the business phone, and point its webhook at n8n. There is no Meta review step, so the setup is short.

Cloud API: more steps, all documented by Meta.

1. A Meta business portfolio ("You must have a business portfolio to use the platform") and a WhatsApp Business Account.
2. A phone number. "Numbers already in use with WhatsApp cannot be registered unless they are deleted first" (source).
3. Or coexistence, if you want to keep the WhatsApp Business app on the same number. It needs app version 2.24.17 or later, runs only through Embedded Signup with a Meta partner, has a fixed throughput of 20 messages per second, and does not support groups, broadcast lists, disappearing or view-once messages, or live location (source).
4. Business verification. It is one route to higher messaging limits, which run 250, 2,000, 10,000, 100,000 and unlimited unique users per rolling 24 hours outside a service window. Limits are shared by every number in the business portfolio.

If you read older guides: the self-hosted On-Premises API is gone. Its final version expired on October 23, 2025.

Does Meta allow AI agents on WhatsApp?

For business agents, yes, with one rule to know. Section 4.7 of the Meta Terms for WhatsApp Business Platform says AI providers, including "general-purpose artificial intelligence assistants", are "strictly prohibited from accessing or using the WhatsApp Business Platform ... when such technologies are the primary (rather than incidental or ancillary) functionality being made available."

A restaurant's ordering agent or a clinic's booking agent uses AI to serve the business's own customers. Our reading is that this is ancillary, not a general-purpose assistant being offered. If your product is a chatbot that answers anything, this clause is aimed at you. Add the escalation-path rule above, and a human handoff is effectively required on the Cloud API anyway.

When should you use Evolution API and when the Cloud API?

Pick the Cloud API when:

• you will message people first: lead reach-back after a form fill, reminders, follow-ups, order updates
• volume will grow, or several numbers need to share one account
• losing the number would hurt the business
• you want the official business profile and verification
• you are in a regulated field

Evolution API can make sense when:

• the agent only replies to customers who message first
• staff must keep using the normal WhatsApp app on the same number, and coexistence's limits (no groups, for example) do not fit
• you want a staff alert group on WhatsApp itself
• budget is tight and Meta fees on templates would matter
• you are prototyping and plan to move to the Cloud API before scaling

These two lists are how we choose. Our restaurant clients fit the second list: reply-only ordering and support, staff on the phone, a WhatsApp escalation group. For lead reach-back and reminders we recommend the Cloud API, because those flows are exactly what WhatsApp's terms and Meta's template rules are written for.

The agent itself barely changes between the two. In n8n, only the receive and send nodes differ, which is why we keep them at the edges of the workflow. Our build log shows the rest, and what drives the price covers the budget side.

Frequently asked questions

Related reading

• Case studyWhatsApp AI Agent for Restaurants: Two Live n8n BuildsRead it
• Articlen8n WhatsApp AI Agent: How We Build Them for ProductionRead it
• ArticleWhatsApp Chatbot Cost: What Drives the Price of an AI AgentRead it