idleprocesscc

idleprocesscc / co-reading-mcp

Public

A local co-reading MCP server for chunked books, reading progress, search, and margin annotations.

11
4
69% credibility
Found May 24, 2026 at 12 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
JavaScript
AI Summary

Co-Reading MCP is a local reading companion that creates a shared space where you and an AI can read books together. You import EPUB or text files into a personal library, read through them section by section while tracking your progress, highlight passages and leave margin notes, and submit your annotations to get thoughtful AI responses. The system preserves chapter boundaries, remembers where you left off, and celebrates when you finish a book. Everything stays stored privately on your own computer.

How It Works

1
📚 You want to read with an AI companion

You discover a tool that lets you and Claude read the same book together, leaving notes in the margins just like a study partner.

2
⚙️ You start the reading room

With a simple command, you launch a local server that becomes your private reading space where books and notes are stored safely on your own computer.

3
📖 You add books to your library

You import EPUB or text files, and the system automatically breaks them into readable sections while keeping chapter boundaries intact.

4
🔖 You read section by section

You move through your book one chunk at a time, with the system remembering exactly where you left off so you can pick up right where you stopped.

5
✏️ You highlight and leave notes

You select any passage and write a margin note attached to it, building up a collection of thoughts and reactions as you read.

6
You discuss with your AI reading partner
🤖
Claude responds to your notes

The AI reads your annotations and attached passages, then adds thoughtful replies under your notes in the margins.

📝
You keep reading and annotating

You continue through the book, adding more notes and reading at your own pace without interrupting your flow.

🎉 You finish your book together

When you complete the last section, the system celebrates your achievement with a special message and a summary of your reading journey and annotations.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 12 to 11 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is co-reading-mcp?

Co-reading-mcp is a local MCP server that turns your AI assistant into a reading companion. Instead of dumping a book into a chat and asking for a summary, you read through content together chunk by chunk, leaving anchored margin notes along the way. The server handles EPUB and plain text imports, breaks them into stable sections, and tracks where you left off. Built in JavaScript with Node.js 18+, it speaks the Model Context Protocol over stdio for local Claude Desktop integration or SSE/HTTP for remote deployments. You get a minimal web reader UI out of the box, but the REST API means any client can hook into the same backend.

Why is it gaining traction?

The hook is the annotation flow. You highlight a passage, scribble a note, stage several notes, then hit "Send to Claude" where context is included once per session without drowning the model in repeated text. When you finish a book, the server returns a small "finish ritual" prompt instead of just marking done. The chunk-based architecture means you can resume exactly where you stopped across sessions. For teams or individuals who annotate heavily while reading, this fills a gap that one-shot summarization tools simply ignore.

Who should use this?

This targets developers who read technical books or documentation with AI assistance and want persistent, anchored notes rather than ephemeral chat memory. Researchers annotating papers, technical book club members coordinating with Claude, or solo devs who want to revisit marginalia from past reading sessions will get the most value. If you just want to summarize a PDF quickly, look elsewhere. If you want a durable reading surface where human and AI both leave marks, this is purpose-built for that.

Verdict

Co-reading-mcp is a focused, thoughtfully designed tool solving a real niche problem. The documentation quality is high, the feature set is coherent, and the architecture respects privacy by keeping data local. However, with only 11 stars and a credibility score of 0.70%, it is early software with untested real-world scale. The web reader is intentionally minimal by design, not omission, but production use will require living with that constraint. Worth trying if the workflow matches your needs, but treat it as a personal tool rather than mission-critical infrastructure.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.