Most ways to script WhatsApp route your data through a paid Business API and a third-party server. This takes the opposite path: it authenticates as your real account by scanning a QR code, then keeps everything on your own machine. The result is that an AI assistant can act inside your actual chat history — the same threads you read on your phone — without that history ever touching a cloud you don't control.
What Sets It Apart
- Two-process split: a Go bridge using the whatsmeow library handles the WhatsApp web multidevice protocol, while a Python MCP server exposes clean tools to the model. That separation means the messy session-management lives away from the LLM-facing surface.
- Full message history lands in local SQLite, so search and retrieval run against your own data rather than paginated remote API calls — and the model only sees what a specific tool call returns.
- It rides the personal-account web API, not the Business API, so there is no per-message billing, no Meta app review, and no phone-number provisioning to get started.
Who It's For
Great fit if you want an AI agent that can triage, summarize, or reply within your own WhatsApp without handing your messages to a SaaS. Look elsewhere if you need official, terms-of-service-blessed automation for a business: this drives a personal account through an unofficial path, the QR session can expire and need re-linking, and running another client against your number carries the usual risk that personal-API usage is not something WhatsApp formally supports.
