The problem with Large Language Model exports isn't storage; it's readability. A developer has released Recto, a local desktop application designed to convert the raw JSON and JSONL logs from ChatGPT and Claude into a structured, searchable book format. While users can easily back up their conversation history, the resulting files are often an unreadable wall of text, making it nearly impossible to retrieve specific decisions or debugging trails from past sessions.

Solving the 'Unreadable Data' Problem

Recto addresses a critical friction point in AI workflows: the disconnect between data ownership and data utility. The tool ingests exports from ChatGPT, Claude, local Claude Code history, and even Doubao or Kimi via bookmarklet. It then renders these logs with chapters, titles, a table of contents, and full-text search capabilities. Users can highlight text in four colors and attach notes, effectively turning a chaotic log file into a curated knowledge base.

Technical Constraints and Local-First Design

The application is built with a strict 'local-first' philosophy. It runs as a small Python process serving a local page to the browser, requiring no account creation, telemetry, or data uploads. The backend relies exclusively on the Python standard library, avoiding third-party dependencies to ensure compatibility across different Windows environments. Currently, the tool is Windows-only, with no macOS build available, and it does not function as a chat client but rather as a post-hoc analysis tool.

Licensing and Pricing Structure

Recto employs an offline licensing model using Ed25519 signatures, meaning activation doesn't require a connection to a central server. The core featuresβ€”reading, importing, highlighting, and searchβ€”are free permanently. AI-assisted features, such as automatic chapterization, require a license or API key. The launch price is set at $9.99 until December 31, 2026, or the first 100 paid licenses, after which the price increases to $14.99 for a one-time purchase.

Key Takeaways

  • Recto transforms raw LLM export files into searchable, annotated books, solving the readability crisis in AI data ownership.
  • The tool is Windows-only and local-first, with no data uploads or accounts required for basic functionality.
  • Core reading and annotation features are free, while AI-enhanced organization features require a one-time purchase.
  • The developer explicitly removed page numbers in favor of stable content anchoring to handle layout reflow issues.

The Bottom Line

This is a necessary tool for anyone who treats AI conversations as serious work products rather than ephemeral chats. While the lack of macOS support is a significant limitation, the focus on local privacy and structured retrieval over raw data dumps is exactly what the ecosystem needs right now.