A developer has launched Coffeetable, a Claude connector that surfaces book pages directly in chat—so users can preview actual text before committing to a purchase. The project landed on Hacker News with the pitch: "What better way to decide you like a book than actually read a few pages." It's a deceptively simple idea that speaks to how people are already using AI assistants as de facto recommendation engines.

How It Works

Coffeetable integrates with Claude as a connector, allowing the model to pull and display excerpts from books within the conversation. Users who've been asking things like "should I read this book?" or "is this worth my time?" can now get their answer by actually reading—not just skimming reviews or summaries. The developer notes that readers already turn to Claude before buying; Coffeetable simply closes the loop between discovery and preview.

Why Book Discovery Matters for AI Assistants

Book recommendation is a natural use case for LLMs. Users trust these models to synthesize opinions, summarize plots, and compare titles—but they still have to leave the chat to actually sample prose style or writing quality. Coffeetable attacks that friction point directly. If you've ever wondered whether an author's voice clicks with you, a quick page preview inside Claude beats tabbing over to Goodreads or Amazon.

The Connector Approach

By building as a Claude connector rather than a standalone app, Coffeetable inherits the model's existing context and conversation history. Your book-related questions stay in one thread; no separate reader interface required. This also means Claude can theoretically recommend books based on what it already knows about your preferences—and then let you preview those recommendations without breaking flow.

Key Takeaways

  • Coffeetable solves the "preview gap" between AI book recommendations and actual purchase decisions
  • The connector architecture keeps discovery conversational rather than fragmenting the experience across apps
  • This pattern—pulling rich media into chat—could extend to other domains like code samples, product specs, or academic papers

The Bottom Line

Coffeetable won't set the HN front page on fire with its score of 4—but it's onto something. The future of AI assistants isn't just answering questions; it's becoming the layer where discovery and action meet. Book previews are a small proof-of-concept, but the underlying pattern has legs.