news-agents
📰 Building News Agents to Summarize News with MCP, Q, and tmux
claude mcp add --transport stdio eugeneyan-news-agents python -m mcp dev src/main.p \ --env CONFIG_DIR="context/" \ --env FEEDS_FILE="feeds.txt (path to feeds list)" \ --env OUTPUT_DIR="summaries/"
How to use
News Agents is an MCP-based terminal workflow that orchestrates multiple agents to fetch, parse, and summarize news articles from sources such as Hacker News, TechCrunch, and the Wall Street Journal. The system uses MCP to expose feeds as tools that the main agent can orchestrate, with several sub-agents operating in parallel to process different subsets of feeds. You interact with the system through the MCP CLI (via uv) and monitor progress in your terminal, often using tmux panes for visibility. To run it, you typically invoke the MCP runner to start the dev workflow, which spins up the main agent and its sub-agents that pull feeds, parse articles, generate summaries, and save them under summaries/.
How to install
Prerequisites:
- Python 3.8+ installed on your system
- uv (the Ultralight Virtualization runtime) installed and available in PATH
- git
Installation steps:
-
Clone the repository: git clone https://github.com/eugeneyan/news-agents.git
-
Change into the project directory: cd news-agents
-
Install and verify MCP tooling (via uv): uv sync uv tree
-
Ensure dependencies for the provided Python MCP script are available. The workflow is designed to be executed with the MCP dev runner, e.g.: uv run mcp dev src/main.p
Note: The workflow relies on the feeds.txt file and the context/ directory for agent prompts and system context. Configure feeds.txt with your target sources and ensure context/main-agent.md (and any additional context files) are present under context/ before starting.
Additional notes
Tips and notes:
- The system uses tmux panes to visualize sub-agents concurrently. If you don’t use tmux, you can still run the main agent and collect summaries, but you won’t have the parallel pane view.
- The main agent reads feeds.txt and splits the feed URLs into 3 chunks for parallel processing by sub-agents. Ensure feeds.txt contains valid, accessible RSS/Atom feed URLs.
- Summaries are written to summaries/[feed-name].md and a combined main-summary.md for an overview.
- If you encounter issues, check uv run mcp dev src/main.p for verbose output and verify that the required context files exist under context/. Common problems include missing feeds.txt, missing context instructions, or a misconfigured MCP runner path.
- Environment variables like FEEDS_FILE, OUTPUT_DIR, and CONFIG_DIR can be used to customize locations; update them to suit your environment.
Related MCP Servers
PPTAgent
An Agentic Framework for Reflective PowerPoint Generation
sandbox
All-in-One Sandbox for AI Agents that combines Browser, Shell, File, MCP and VSCode Server in a single Docker container.
FireRed-OpenStoryline
FireRed-OpenStoryline is an AI video editing agent that transforms manual editing into intention-driven directing through natural language interaction, LLM-powered planning, and precise tool orchestration. It facilitates transparent, human-in-the-loop creation with reusable Style Skills for consistent, professional storytelling.
evo-ai
Evo AI is an open-source platform for creating and managing AI agents, enabling integration with different AI models and services.
skillz
An MCP server for loading skills (shim for non-claude clients).
TradingAgents mode
TradingAgents-MCPmode 是一个创新的多智能体交易分析系统,集成了 Model Context Protocol (MCP) 工具,实现了智能化的股票分析和交易决策流程。系统通过多个专业化智能体的协作,提供全面的市场分析、投资建议和风险管理。