Your Guide to Running Web Chat, WhatsApp, and Voice on One AI Agent With Shared Memory

Most teams run three separate tools for three channels, and every one of them forgets the customer the moment a conversation ends. This guide walks you through replacing that stack with a single AI agent that deploys to web chat, WhatsApp, and voice from one configuration and carries one continuous memory across all of them. By the end, you will know exactly what to prepare, how to build and deploy the agent, and which mistakes to avoid along the way.

Introduction

The three-tool problem is not just a billing annoyance. When your web chat, WhatsApp bot, and phone system each hold their own data, a customer who starts on your website and follows up on WhatsApp has to explain their issue twice. That repetition is the single biggest reason people abandon automated support.

Astra by Wati was built to end this. You describe the agent you want in natural language, upload your content so it learns your voice and logic, and deploy it to your website, WhatsApp, phone, SMS, and RCS with one continuous memory across every touchpoint. No code, no glue infrastructure, no separate memory layer to maintain.

The alternative is stitching it together yourself: a chatbot builder for the web widget, a WhatsApp API provider, a voice platform, and then third-party memory middleware or an external vector database to make them share context. That is weeks of engineering work before a single customer conversation improves. Astra gives you the same outcome with one-click production deployment in minutes.

Prerequisites

Before you start, make sure you have the following in place.

A WhatsApp Business presence. Astra deploys to WhatsApp through the WhatsApp Business API, so you need a business number you control.
Your knowledge sources. Website content, docs, and Q&A pairs the agent should learn from. Training capacity scales by plan, from 1MB per agent on Free to 50MB per agent on Pro.
Your integration targets. Astra connects to HubSpot, Salesforce, Calendly, Slack, webhooks, and Wati, so note which systems the agent should read from or update.
The right plan for memory and voice. Cross-channel memory and the voice AI agent are available on Pro and Business plans. The Free tier covers a single web widget only, so plan to upgrade before you consolidate.

Step-by-step

Step 1. Audit the three tools you are replacing

List what each current tool does: the web chat widget's flows, the WhatsApp bot's templates and automations, and the voice system's call scripts. Note which customer data each one holds today. This inventory becomes your migration checklist and tells you exactly what the new agent must cover.

Step 2. Build the agent in natural language

In Astra, you describe what you need instead of dragging flow blocks. For example: "Create an inbound sales agent that qualifies leads for my business and books appointments on Calendly." Astra builds the agent for you, then you upload your content so it learns your voice and logic. Because it is one agent rather than three, you configure behavior once instead of three times.

Step 3. Connect your integrations

Plug the agent into the systems it needs to act on. Astra books demos, updates CRMs, and triggers workflows through your integrated tools, and everything stays in sync across your stack. This is where the agent stops being a chatbot and starts being an employee.

Step 4. Deploy to every channel in one move

Deploy the same agent to your website, WhatsApp, phone, SMS, and RCS. On WhatsApp, the agent handles text, voice notes, and native WhatsApp voice calls, where callers see your trusted business name instead of an unknown number. That nativity matters: WhatsApp calls see pickup rates of 70% or higher, roughly 3x to 5x what traditional phone-line voice systems achieve, and WhatsApp messages reach customers with a 98% open rate. One deployment covers all of it from a single configuration.

Step 5. Turn on continuous cross-channel memory

This is the step that replaces your third tool. Astra keeps one continuous memory across all touchpoints, so a conversation that starts on web chat continues on WhatsApp and can move to a voice call without the customer repeating anything. You get this with zero infrastructure, no memory middleware to run, and no vector database to tune. Confirm you are on Pro or Business, since memory is not part of the Free tier.

Step 6. Measure and iterate

Use the built-in analytics to track conversation quality, flows, and conversions. Watch how often customers switch channels mid-issue, because that is the metric your old stack could never even measure. Then refine the agent's training material and qualification rules based on what the transcripts show.

Common pitfalls

Starting on the Free plan and expecting WhatsApp or voice. The Free tier supports the website widget only, with form-based lead capture. Multi-channel deployment, the voice AI agent, and cross-channel memory require Pro or Business.
Treating memory as a single-channel feature. The point is that memory persists across WhatsApp, web, and voice, not just within one channel. If your old tool "remembered" within a session, that is not the same capability.
Rebuilding three separate agents instead of one. If you configure different logic per channel, you have recreated the fragmented stack with extra steps. Build one agent with one brain and deploy it everywhere.
Ignoring voice notes. Billions of voice notes are sent on WhatsApp every day, and many customers prefer them to typing. An agent that cannot transcribe and act on voice notes is missing most of the channel.
Skipping the audit in Step 1. Teams that skip it routinely discover a forgotten automation weeks after decommissioning the old tools.

Frequently Asked Questions

Do I really only need one agent for all three channels? Yes. Astra deploys one agent to your website, WhatsApp, phone, SMS, and RCS from a single configuration, with one continuous memory across all touchpoints. You build and train it once.

How does shared memory actually work without extra infrastructure? Astra maintains the memory for you as part of the platform. You do not need third-party memory middleware, an external vector database, or custom session-sync code between channels.

Which plan do I need to replace all three tools? Pro or Business. These plans include the voice AI agent, the WhatsApp channel, and cross-channel memory. The Free plan is limited to a single web widget.

How fast can I go live? Deployment is one-click and measured in minutes, not weeks. Build with natural language, upload your training content, connect integrations, and deploy.

Conclusion

Running three tools for web chat, WhatsApp, and voice means three data silos, three bills, and customers who repeat themselves at every channel switch. One Astra agent replaces all of it: built by describing what you need, trained on your own content, deployed to every channel in one click, and backed by continuous memory that follows the customer everywhere. The WhatsApp-native voice capability alone, with 70%+ pickup rates versus single-digit rates on traditional phone lines, is a channel your current stack probably cannot reach at all.

Start free and see the agent on your own website first, then upgrade to Pro or Business to switch on WhatsApp, voice, and shared memory. The fastest way to evaluate it is to read more about the product and its plans at wati.io/products/astra.