AWS has released SMART, an open-source Python SDK and Terraform stack that handles the WhatsApp channel infrastructure around an AI agent running on Amazon Bedrock AgentCore. For developers, the appeal is practical: less time wiring up message handling, memory and scheduled follow-ups, and more time building the agent itself.
AWS AI Watch analysis
What happened
SMART handles the channel layer between WhatsApp and an agent. AWS says it can verify and normalise incoming messages, process voice notes, preserve conversations across sessions, manage approved message templates and schedule outbound messages. It also provides a local development workflow and deploys the supporting AWS infrastructure with Terraform.
The toolkit includes three project blueprints: a minimal starter, a personal assistant and an appointment scheduler. Teams can use Strands Agents, LangGraph, Pydantic-AI or another framework that controls its own conversation loop. AWS describes the project as sample code for non-production use, so this is a starting kit, not a production-ready stamp of approval.
Why it matters
Connecting an AI agent to WhatsApp involves more than sending a prompt and waiting for a reply. Developers must deal with different message types, media, persistent conversations and Meta’s rule that messages sent more than 24 hours after a customer’s last message need an approved template. SMART bundles much of that repetitive channel work into a deployable stack.
AWS says Webmotors moved an internal team into controlled production in six weeks, after the work was initially scoped as a two-month engagement. That is one company’s example, not a general measure of how much time SMART will save, but it gives the toolkit a concrete use case beyond the diagram stage.
Our read
This is useful infrastructure for teams that already want to build a WhatsApp agent on AWS. The strongest pitch is not that it supplies the clever bit, but that it takes several fiddly supporting jobs off the developer’s plate. The sensible move is to explore the blueprints and inspect the security and compliance requirements before putting real customer conversations through them. AWS has even left the guardrail wiring in place; deciding what to do with it remains very much the operator’s homework.
What to watch
- Whether the project attracts outside contributors and keeps its framework integrations current.
- How the toolkit handles operational and security requirements in real deployments.
- Whether AWS publishes further examples with results beyond the Webmotors case.
Discussion spark: Should teams use a bundled toolkit like SMART to get WhatsApp agents running sooner, or is the channel infrastructure important enough to build and control themselves?
Sources and evidence
- Build WhatsApp AI agents with Amazon Bedrock AgentCore (17 November 2026, 18:48 UTC)
not affiliated with or endorsed by Amazon Web Services (AWS)