The best WhatsApp chatbot builder for your business depends on the conversation you need to finish. Collecting a lead, answering a policy question and changing an authenticated customer's order require different workflows. A visual editor alone does not establish that all three will work.
This guide compares five approaches: Manychat, LoopReply, Respond.io, Wati and Twilio. Some provide packaged automation and an inbox; others require more integration work. Start with one customer journey, then compare the setup, operator experience and complete bill for that journey.
Reviewed October 4, 2026. This comparison draws on current official documentation and LoopReply's configuration implementation, rather than a live-account benchmark. We publish LoopReply. Use the linked sources and workflow checks below to evaluate which option fits your business.
Quick comparison: what you need to evaluate
| Option | Documented starting point | Cost items to check | First demonstration to request |
|---|---|---|---|
| Manychat | WhatsApp automations and social messaging journeys | Applicable account plan, contact allowance, AI eligibility and WhatsApp message charges | A lead enters a flow, asks an unexpected question and reaches a person |
| LoopReply | Conversational Agent configuration, knowledge and workflows; a Meta Cloud API connection requiring setup | Current subscription, usage limits and separate messaging costs | Your configured phone number routes to the intended Agent and an operator can complete the conversation |
| Respond.io | Team messaging operations with AI Agents, knowledge and human takeover | Monthly Active Contacts, users, AI credits and separate WhatsApp fees | AI qualification followed by a handover to the correct team |
| Wati | WhatsApp campaigns, automation and team inbox, with additional channels and AI options | Selected plan, messaging rates, users and separately priced Astra where applicable | An approved campaign or support flow followed by operator handling |
| Twilio | Messaging APIs and developer tooling | Per-message handling, Meta charges, other products and implementation work | A test sender receives a reply from your application, including its failure path |
The vendor sections below link the underlying sources and explain the setup questions to resolve.
WhatsApp pricing: count messages, not just conversations
Meta's Business Platform pricing page describes charging on delivered messages, with rates determined by category and destination market. A monthly chatbot subscription is a separate cost.
The 24-hour customer service window is also a messaging rule, not a guarantee that your total bill is zero. Confirm which messages require approved templates and which categories, allowances and provider fees apply to your account.
Watch the October 2026 change. Manychat's WhatsApp pricing guide, updated September 30, says service messages become chargeable on October 1 after a monthly allowance of 1,000 service messages per business phone number. It publishes separate rate tables for dates before and after that change. That is Manychat's documented billing guidance; use the applicable provider's current terms for your own account.
During this review, general Meta overview wording and parts of provider pricing pages still differed on service-window charges. Use the effective-date rate card, recipient market and actual provider quote when budgeting. Avoid treating an older “free service conversations” statement as a current cost estimate.
1. Manychat: evaluate the social-to-WhatsApp journey
Manychat's WhatsApp product page describes FAQ automation, lead qualification and follow-up. Its AI product page also advertises AI replies and conversation tools. Verify the specific AI feature on WhatsApp and on your plan before including it in your workflow.
Ask to see the whole journey: entry point, qualification questions, unexpected replies, operator access and the next action. This matters more than whether the demo's opening message sounds natural.
Pricing needs an account-specific check. Manychat's billing FAQ distinguishes plans introduced on March 2, 2026 from older accounts and says availability varies by region. Do not apply an old $15 starting price to every new account. Confirm the current subscription and contact calculation, then add the WhatsApp charges described above.
For the broader website-versus-social decision, see our LoopReply and Manychat comparison.
2. LoopReply: inspect the connection as well as the Agent
LoopReply provides conversational Agent configuration, knowledge sources and workflows. Its WhatsApp configuration uses the Meta Cloud API. The connection form asks for an App ID, App Secret, access token, Phone Number ID and verification token, with an optional Business Account ID. It also provides a Workspace webhook URL to configure in the Meta app.
This involves Meta account and webhook setup. Saving those fields is not proof that messages are arriving, that the right Agent is assigned or that an operator can reply. Before choosing LoopReply for a WhatsApp deployment, test those steps with a phone number and recipients you control.
Evaluate the Agent separately: supply relevant knowledge, ask a known question, ask one the sources cannot answer and request a human. Then repeat on WhatsApp. A successful website preview does not prove the channel connection, media behavior, templates or outbound messaging rules.
In Architect Studio, Agent instructions, appearance and workflow changes use drafts and publishing. Knowledge and model-context settings such as response style or language are live settings. Check the field being changed before assuming every experiment is isolated from customers.
Confirm the current LoopReply plan and allowance, WhatsApp availability for the intended setup and separate messaging charges. We do not present a free website Agent allowance as a promise of free WhatsApp operation.
3. Respond.io: compare AI and team operations together
Respond.io's AI Agents page describes Agents that use approved knowledge, perform configured actions and escalate to humans.
For a team inbox, test the transition between automation and an operator. Can the person see the relevant context, take ownership and prevent overlapping replies? If the Agent updates a CRM, demonstrate a failed update as well as a successful one.
Respond.io's pricing page separates Monthly Active Contacts, users and AI credits, and states that WhatsApp fees are additional. The page also describes AI-credit overages, alongside broader included-AI language. Ask for the exact included allowance and overage behavior on the plan quoted to you. An “AI included” label should not replace that calculation.
4. Wati: check the plan and AI package you are buying
Wati's current pricing and feature page lists campaign automation, team inbox features and channels beyond WhatsApp, including Instagram. It also describes Astra AI Agents as a separately priced add-on in relevant plan descriptions.
For a campaign-led use case, demonstrate contact permission, template selection, delivery handling and an operator follow-up. For AI support, ask which knowledge and action features belong to the selected Wati plan and which require Astra or another add-on.
Request a quote with the currency, billing interval, user count, message charges and any AI package shown separately. A headline subscription price does not answer whether the complete workflow fits the budget.
5. Twilio: useful when someone owns the implementation
Twilio provides messaging infrastructure. Its WhatsApp quickstart walks through sending and receiving messages, which is a different starting point from buying a finished support inbox. Decide who will build and maintain the application, AI logic, knowledge access and operator workflow around that connection.
On the review date, Twilio's WhatsApp pricing page listed a $0.005 handling fee per inbound or outbound message, alongside separate Meta charges. Other products, add-ons and applicable fees can affect the total.
As a simple hypothetical calculation, 2,000 messages at that handling rate equal $10 in Twilio handling fees alone. That is not a total monthly chatbot price. Add the relevant Meta charges, application hosting, AI usage, other purchased services and maintenance before comparing it with a packaged tool.
A cost worksheet for any market
Use the same workload for every vendor. A US business messaging UK customers may face different destination-market charges from a UK business messaging US customers. “Global English” describes an audience, not a single WhatsApp rate.
| Input | What to record |
|---|---|
| Recipient markets | Countries or market groups, with expected message volume for each |
| Message categories | Marketing, utility, authentication and service, including eligibility restrictions |
| Billable volume | Delivered outbound messages by category, applicable allowances and provider handling of inbound messages |
| Platform subscription | Plan, currency, billing interval and tax treatment |
| Contact and operator usage | Included contacts, users and expected overages |
| AI and connected systems | Included credits, overages, model/API costs and paid integrations |
| Operational work | Setup, template maintenance, knowledge updates, monitoring and human follow-up |
Ask for both a normal month and a busier month. Keep the assumptions visible. Do not mix contacts, messages, conversations and resolved cases as though they were the same billing unit.
Test one workflow before committing
Run these checks with your own account and test recipients:
- Confirm account and number eligibility. Check ownership, verification, existing-number migration or coexistence requirements with the chosen provider. Avoid assuming every existing WhatsApp number can move without interruption.
- Define the permitted conversation. Record where users enter, what they agreed to receive and how they can stop unwanted follow-up. Review the applicable platform requirements for the intended message type.
- Try a known answer and a missing answer. Use a small, current reference set. The missing-answer test should lead to clarification or a useful next step, not invented policy.
- Test a private-data request. An order number or phone number alone should not be treated as sufficient proof that someone may see a record. Demonstrate the identity and permission check your business requires.
- Ask for a person. Check assignment, context, reply ownership and what the customer sees when no operator is available.
- Test the window and failure path. Include an attempted follow-up outside the service window, rejected templates, expired credentials and a failed connected-system request. Confirm what the operator can see and do.
- Read the usage record. Reconcile the messages sent during the test with the provider's billable units before extrapolating a monthly cost.
WhatsApp bots for construction and field service
A useful first construction WhatsApp bot workflow is an intake request that a person can act on. Ask for the existing job reference, if there is one, the site postcode or service area, the work or issue needing attention, relevant access details and the customer's preferred contact time. Collect only what the team needs for that request.
Show the customer a short summary to correct, then pass it to the responsible person. Confirm receipt and explain that a team member will review the request and agree the next step. Test that the person actually receives the details and can reply before making that confirmation part of the workflow.
A preferred time is a request, not a confirmed appointment. Only promise a quote, booking or arrival time after the responsible person or connected system confirms it. If the channel accepts photos, treat them as supporting information for review; receiving an image does not establish that the bot can assess a fault.
Common buying questions
Can I build a WhatsApp chatbot without coding?
Some products provide visual automation, but the amount of external setup varies. Confirm who configures the business account, phone number, templates, data connections and human handover. Twilio is an infrastructure option here, not a claim of the same setup experience as a packaged builder.
Is a free chatbot plan enough for WhatsApp?
Not necessarily. A free website plan or trial does not establish WhatsApp channel eligibility, messaging credits or the ongoing provider bill. Check the account-specific plan and effective-date rate card.
Does an AI Agent automatically know my order and booking data?
No. Supplying public FAQs is different from authenticating a customer, retrieving a private record or making a booking. Request a demonstration of the exact integration and its permissions. Start with our guide to using your own data to separate knowledge setup from business actions.
Which builder should a small business choose?
Choose the one that completes your required journey, fits the team's operating skills and has a cost you can explain from its billing units. Use the same questions and workload across your shortlist. Our AI chatbot buyer's guide provides a broader evaluation worksheet before you commit.
Match the shortlist to one entry point
For a website visitor who needs a product answer and then a team reply, evaluate the knowledge source and receiving inbox first. For a social campaign that leads into WhatsApp, evaluate the campaign trigger and opt-in journey first. For a custom service operation, include the developer who will maintain the connection. This is a fit-based shortlist, not a claim of measured winner rankings.
Run one fictional project enquiry through the chosen channel: service area, requested work, preferred follow-up, then a request for a person. Confirm both inbound delivery and the operator's outbound reply. Review the approved WhatsApp account, sender, credentials, webhook, and current provider message rules before treating a connected label as a completed installation.
Use the Manychat comparison for website-versus-social fit and the construction enquiry example for a bounded field-service flow. The business buyer guide helps compare the full workload rather than the cheapest entry plan.