New Claude Code Skill Recovers Export-Blocked Kindle Highlights on macOS
Developer l3a0 uses local OCR and SQLite databases to bypass undocumented Amazon export limits.
A new Claude Code skill is allowing macOS users to reclaim personal reading data that Amazon routinely hides or truncates during export. The tool targets the undocumented limits of the Kindle notebook, ensuring users can recover their own curated highlights in full.
Developed by l3a0, the tool specifically recovers content that triggers the Amazon notice: "Some highlights have been hidden or truncated due to export limits." To achieve this, the skill orchestrates a multi-step local workflow. It extracts precise character positions from the Mac Kindle app's SQLite database, takes canvas captures of the Cloud Reader, and processes those images using local OCR via Apple's Vision framework. The final output is a verbatim, location-cited Markdown file.
The Export Gap
Amazon imposes undocumented restrictions on the Kindle notebook page (read.amazon.com/notebook). These limits often prevent users from exporting their complete set of notes to external formats, effectively locking user-generated data within a closed ecosystem. While the web interface enforces these caps, the local Mac Kindle application maintains a more comprehensive record of annotation positions in its internal database, providing a loophole for data recovery.
Why It Matters
This project highlights the utility of the Model Context Protocol (MCP) and Claude Code in managing complex, on-device operations. By combining browser automation, local database access, and hardware-accelerated OCR, the tool reclaims user data without attempting to bypass digital rights management (DRM). As developer l3a0 noted, these are the user's own notes in their own account, yet they are capped by a budget that Amazon neither documents nor allows users to increase.
Technical Constraints and Results
Because the tool relies on AppleScript, Apple Vision, and specific local data files from the Mac Kindle app, it is strictly limited to macOS. In testing across four different books, the tool's efficacy was absolute: out of 2,432 highlights, 815 were found to be export-blocked by Amazon, and all 815 were successfully recovered.
Users looking to implement the tool can find the implementation in l3a0's claude-plugins repository on GitHub. Future developments may focus on expanding these recovery capabilities to other closed-ecosystem note-taking tools.