Abstract
We present Paperboy, a research paper discovery system designed for AI practitioners who build products, not publish papers. Unlike existing tools optimized for academic citation graphs, Paperboy filters the daily arXiv firehose to surface only papers relevant to your active projects—then translates theoretical contributions into actionable implementation guidance. Use Paperboy directly via MCP in Claude or ChatGPT, or set up our n8n automation with Arca to receive matched papers in your inbox.
1. Introduction
1.1 The Problem
You build AI products. You don't have time to read papers.
Every day, arXiv publishes 100+ AI papers. Somewhere in that flood is a technique that could improve your retrieval accuracy, fix your agent's memory problem, or obsolete what you're currently building.
You'll never find it.
1.2 Why Existing Tools Fail
- arXiv — Built for researchers browsing by category, not builders searching by problem
- Semantic Scholar — Optimized for citations and academic influence, not practical implementation
- Twitter/X — Algorithm-driven hype, not relevance to your work
- Newsletters — Generic summaries for everyone, personalized for no one
The result: You've stopped trying. Papers happen to other people.
2. Methods
2.1 MCP-Native Architecture
Paperboy is an MCP server. Connect it to Claude, ChatGPT, or any MCP-compatible client:
mcp.paperboy.sh
Three tools:
| Tool | Function |
|---|---|
get_recent_papers | Browse latest AI papers with filtering |
search_papers | Semantic search across the corpus |
dig_deeper_into_paper | Full text extraction with metadata |
2.2 Personalized Matching with Arca
For papers that come to you—matched against your actual projects:
Paperboy + Arca = Personalized research intelligence
What is Arca? Arca is a private data vault designed for AI agent use. Store descriptions of your current AI projects there via the Arca MCP server (works with Claude, ChatGPT, and n8n). The n8n automation then retrieves your project context from Arca and matches incoming papers against your work to score relevance.
- Sign up at paperboy.sh — Get your free API key
- Store your AI projects in Arca — Use the Arca MCP server to save descriptions of what you're building
- Import our n8n workflow template — This automation uses both Paperboy MCP and Arca MCP to match papers to your projects
- Receive matched papers in your inbox — Only what's relevant, scored by how well it matches your projects, with implementation recommendations
How it works: The n8n workflow fetches new papers from Paperboy, retrieves your project context from Arca, then uses an AI agent to compare them. You receive only papers that matter—with scores and specific recommendations for how to apply them to your work.
3. Results
3.1 Sample Output
3.2 Before / After
| Before | After |
|---|---|
| 0 papers read | Only relevant papers, summarized |
| Missed breakthroughs | Matched to your projects |
| Hours skimming abstracts | 30-second relevance decisions |
| "I should read more" guilt | Actionable insights in your inbox |
4. Getting Started
Option A: Chat Interface
Add Paperboy to Claude or ChatGPT as an MCP connector:
mcp.paperboy.sh
Option B: Automated Pipeline
- Sign up at paperboy.sh (free)
- Get your API key
- Store your projects in Arca (private data vault for AI agents)
- Download the n8n template and follow setup instructions
- Receive daily matched papers in your inbox