The new open-source Go tool, WhatsKept, allows you to transform encrypted WhatsApp backups from iOS into a structured SQLite database, opening up the possibility of intelligent search through personal correspondence using LLM agents.
What Happened
A tool called WhatsKept has been developed that performs an ETL process to convert encrypted WhatsApp backups (iOS) into a local SQLite database with FTS5 full-text search support. The program extracts messages, images, voice notes, and PDF documents, creating a workspace that can be connected to modern AI agents such as Cursor, Windsurf, or Claude Code. For multimodal tasks, cloud enrichment via OpenRouter is provided, allowing for image descriptions and audio transcription.
Context
The project represents a "plumbing" class solution (infrastructure middleware) that addresses the problem of accessing closed personal data. Instead of modifying the original data or breaking encryption, WhatsKept creates a local index that allows for the integration of private communication history into RAG (Retrieval-Augmented Generation) pipelines without transmitting sensitive information to the cloud in its raw form.
Why It Matters for the Industry
For the AI industry, this is an important step in the development of Personal Knowledge Management tools. The project demonstrates the technical feasibility of creating a secure bridge between closed ecosystems (iOS/WhatsApp) and LLM agents, which could lead to the standardization of middleware solutions for providing personal context to models in AI IDEs and local systems.
Why It Matters for Users
Regular users gain the ability to use familiar AI tools to work with their "digital memory." This allows for quickly finding forgotten details, receipts, important agreements, or instructions simply by asking questions in a chat with an AI agent, instead of manually scrolling through years of message archives.
What Is Not Yet Known / Limitations
There is legal and ethical uncertainty regarding the legitimacy of bypassing encryption for LLM context needs, as well as potential privacy risks when using third-party services like OpenRouter for data enrichment.
Sources
Author
Look at AI, Editorial Staff