io.github.xodn348/readhn

Official

by xodn348 · Python

AI-native HackerNews MCP Server. Find HN content that matters with explainable quality signals.

Find relevant HackerNews content with explainable quality signals for AI agents.

README.md3 KBView on GitHub

readhn

AI-native HackerNews MCP Server. Find HN content that matters with explainable quality signals.

What It Does

Discover — Filter stories by keywords, scores, time. Get ranked results with quality signals.

Trust — Find domain experts. See who's talking and why they matter. EigenTrust propagation from seed experts.

Understand — Every result explains WHY. 5 signals: practitioner depth (30%), thread depth (20%), expert involvement (20%), velocity (15%), references (15%).

Quick Start

bash

# Install
pip install readhn

# Auto-configure supported AI agents
readhn setup

readhn setup detects Claude Code, Codex, Cursor, Claude Desktop, Cline, Windsurf, and OpenCode config paths and adds the readhn MCP server.

Useful setup flags:

bash

readhn setup --list              # Show detected agents
readhn setup --dry-run           # Preview config changes only
readhn setup --agents "Cursor"  # Configure only specific agents

After setup, your AI agent auto-discovers readhn and uses it when you ask HN questions.

Usage

Ask your AI agent:

"Show me top HN stories about Rust this week"
"Find experts who write about databases on HN"
"What did practitioners say about Kubernetes networking?"

The agent calls readhn tools, gets results with quality signals, and explains why each result matters.

Configuration (Optional)

bash

export HN_KEYWORDS="ai,llm,rust,distributed-systems,databases"  # Default filter keywords
export HN_MIN_SCORE="50"                                         # Minimum story score
export HN_EXPERTS="tptacek,simonw,antirez,ept,jepsen"           # Seed experts for trust
export HN_TIME_HOURS="24"                                        # Time window

How It Works

When you ask HN questions, your AI agent uses these tools:

discover_stories() — Top stories filtered by keywords/score/time, ranked by quality signals
search() — Algolia search with explainable ranking
find_experts() — Find domain experts using EigenTrust on comment graph
expert_brief() — User profile + activity + trust score
story_brief() — Story + top comments + signals in one call
thread_analysis() — Full comment tree with quality signals per comment

Every response includes signals breakdown: why each result was chosen.

License

MIT

Install

claude_desktop_config.json

{
  "mcpServers": {
    "readhn": {
      "command": "uvx",
      "args": [
        "readhn"
      ]
    }
  }
}