![\"The](\"https://eugeneyan.com/assets/news-agents.webp\" "\\"The")
To better understand MCPs and agentic workflows, I built news-agents to help me generate a daily news recap. It’s built on Amazon Q CLI and MCP. The former provides the agentic framework and the latter provides news feeds via tools. It also uses tmux to spawn and display each sub-agent’s work. At a high level, here’s how it works:
\nHere, we’ll walk through how the MCP tools are built and how the main agent spawns and monitors sub-agents. Each sub-agent processes its allocated news feeds and generates summaries for each feed. The main agent then combines these summaries into a final summary. Here’s the three-minute 1080p demo (watch till at least 0:30):
\n\nEach news feed has its own rss reader, parser, and formatter. These handle the unique structure and format of each feed. (Perhaps in future we can just use an LLM to parse these large text blobs reliably and cheaply.) For example, here’s the code for fetching and parsing the Hacker News rss feed:
\n\n """\n Fetch Hacker News RSS feed.\n\n Args:\n feed_url: URL of the RSS feed to fetch (defaults to Hacker News)\n """\n = \n \n \n = await \n \n return \n \n return f\n \n return f\n \n return f\n\n\n\n """Parse RSS content into a list of story dictionaries."""\n = \n \n = \n = \n\n \n = \n \n\n return \n \n return \nThe MCP server then imports these parsers and sets up the MCP tools. MCP makes it easy to set up tools with the @mcp.tool() decorator. For example, here’s the Hacker News tool:
# Initialize FastMCP server\n = \n\n\n\n\n """Get top stories from Hacker News.\n\n Args:\n feed_url: URL of the RSS feed to use (default: Hacker News)\n count: Number of stories to return (default: 5)\n """\n = await \n \n return \n\n = \n\n # Limit to requested count\n = \n\n \n return \n\n = \n return \nWith this, here’s the tools that our agent has from both news-mcp and the built-ins:
\n\n- ----\n- ----Next, we’ll set up a multi-agent system to parse the news feeds and generate summaries. We’ll have a main agent (image below, top-left) that coordinates three sub-agents, each running in a separate tmux window (image below; bottom-left and right panes).
\n
The main agent (top-left) with its newly spawned sub-agents
\nThe main agent will first divide the news feeds into three groups. Then, it’ll spawn three sub-agents, each assigned to a group of news feeds. In the demo, these sub-agents are displayed as a separate tmux window for each sub-agent.
\n\n\n- --\n\nTruncated instructions for main agent
\nThen, the sub-agents will process their assigned news feeds and generate summaries for each of them. The sub-agent also categorizes stories within each feed, such as AI/ML, technology, business, etc. Throughout this process, the sub-agents display status updates. When the sub-agent finishes processing its assigned feeds, it displays a final completion status.
\n\n\nTruncated instructions for sub-agent
\nWhile the sub-agents are processing their assigned feeds, the main agent monitors their progress. When all sub-agents are done, the main agent reads the individual feed summaries and combines them into a final summary.
\nFinally, we define the news feeds in feeds.txt below. We have six feeds: Hacker News, The Wall Street Journal Tech, The Wall Street Journal Markets, TechCrunch, AI News, and Wired.
\nAnd here’s the truncated main-summary it generated for 4th May.
\n\n- ---\n\n\n\n\n1. - - -\n2. - - -\n3. - - -\n4. - - -\nTry it for yourself! Install Amazon Q CLI and play with the code here: news-agents.
\n\n\n\n\n\n\nInitially, I wanted to host this as a web app, perhaps on a platform like Daytona. However, I quickly learned that building remote MCPs isn’t trivial, especially with only a couple of weekend hours to hack on this. For now, I’ll explore applying this setup to other use cases such as parsing design docs and COEs, or multi-agent writing and coding workflows. If you’re also experimenting with MCPs or agentic workflows, I’d love to hear from you!
\nIf you found this useful, please cite this write-up as:
\n\n\nYan, Ziyou. (May 2025). Building News Agents for Daily News Recaps with MCP, Q, and tmux. eugeneyan.com. https://eugeneyan.com/writing/news-agents/.
\n
or
\n@article{yan2025news-agents,\n title = {Building News Agents for Daily News Recaps with MCP, Q, and tmux},\n author = {Yan, Ziyou},\n journal = {eugeneyan.com},\n year = {2025},\n month = {May},\n url = {https://eugeneyan.com/writing/news-agents/}\n}\nShare on:
\n![\"\"](\"https://eugeneyan.com/assets/icon-twitter.svg\"){kind=icon}
![\"\"](\"https://eugeneyan.com/assets/icon-linkedin.svg\"){kind=icon}
![\"\"](\"https://eugeneyan.com/assets/bluesky.svg\")
![\"\"](\"https://eugeneyan.com/assets/icon-facebook.svg\"){kind=icon}
![\"\"](\"https://eugeneyan.com/assets/icon-mail.svg\"){kind=icon}
Join 11,800+ readers getting updates on machine learning, RecSys, LLMs, and engineering.
\n", "content_markdown": "To better understand MCPs and agentic workflows, I built news-agents to help me generate a daily news recap. It’s built on [Amazon Q CLI](https://docs.aws.amazon.com/amazonq/latest/qdeveloper-ug/command-line.html) and [MCP](https://modelcontextprotocol.io). The former provides the agentic framework and the latter provides news feeds via tools. It also uses [tmux](https://github.com/tmux/tmux/wiki) to spawn and display each sub-agent’s work. At a high level, here’s how it works:\n\n```markdown\nMain Agent (in the main tmux pane)\n├── Read feeds.txt\n├── Split feeds into 3 chunks\n├── Spawns 3 Sub-Agents (in separate tmux panes)\n│ ├── Sub-Agent #1\n│ │ ├── Process feeds in chunk 1\n│ │ └── Report back when done\n│ ├── Sub-Agent #2\n│ │ ├── Process feeds in chunk 2\n│ │ └── Report back when done\n│ └── Sub-Agent #3\n│ ├── Process feeds in chunk 3\n│ └── Report back when done\n└── Combine everything into main-summary.md\n```\n\nHere, we’ll walk through how the MCP tools are built and how the main agent spawns and monitors sub-agents. Each sub-agent processes its allocated news feeds and generates summaries for each feed. The main agent then combines these summaries into a final summary. Here’s the three-minute 1080p demo (watch till at least 0:30):\n\n[YouTube video player](https://www.youtube.com/watch?v=q41YevguhQw)\n\n## Setting up MCPs for news feeds\n\nEach news feed has its own rss reader, parser, and formatter. These handle the unique structure and format of each feed. (Perhaps in future we can just use an LLM to parse these large text blobs reliably and cheaply.) For example, here’s the code for fetching and parsing the Hacker News rss feed:\n\n```python\nasync def fetch_hn_rss(feed_url: str) -> str:\n \"\"\"\n Fetch Hacker News RSS feed.\n\n Args:\n feed_url: URL of the RSS feed to fetch (defaults to Hacker News)\n \"\"\"\n headers = {\"User-Agent\": USER_AGENT}\n async with httpx.AsyncClient() as client:\n try:\n response = await client.get(feed_url, headers=headers, timeout=10.0)\n response.raise_for_status()\n return response.text\n except httpx.HTTPError as e:\n return f\"HTTP Error fetching RSS: {str(e)}\"\n except httpx.TimeoutException:\n return f\"Timeout fetching RSS from {feed_url}\"\n except Exception as e:\n return f\"Error fetching RSS: {str(e)}\"\n\n\ndef parse_hn_rss(rss_content: str) -> List[Dict[str, Any]]:\n \"\"\"Parse RSS content into a list of story dictionaries.\"\"\"\n stories = []\n try:\n root = ET.fromstring(rss_content)\n items = root.findall(\".//item\")\n\n for item in items:\n story = {\n \"title\": item.find(\"title\").text\n if item.find(\"title\") is not None\n else \"No title\",\n \"link\": item.find(\"link\").text if item.find(\"link\") is not None else \"\",\n \"description\": item.find(\"description\").text\n if item.find(\"description\") is not None\n else \"No description\",\n \"pubDate\": item.find(\"pubDate\").text\n if item.find(\"pubDate\") is not None\n else \"\",\n # Any other fields we want to extract\n }\n stories.append(story)\n\n return stories\n except Exception as e:\n return [{\"error\": f\"Error parsing RSS: {str(e)}\"}]\n```\n\nThe MCP server then imports these parsers and sets up the MCP tools. MCP makes it easy to set up tools with the `@mcp.tool()` decorator. For example, here’s the Hacker News tool:\n\n```python\n# Initialize FastMCP server\nmcp = FastMCP(\"news-mcp\")\n\n\n@mcp.tool()\nasync def get_hackernews_stories(\n feed_url: str = DEFAULT_HN_RSS_URL, count: int = 30\n) -> str:\n \"\"\"Get top stories from Hacker News.\n\n Args:\n feed_url: URL of the RSS feed to use (default: Hacker News)\n count: Number of stories to return (default: 5)\n \"\"\"\n rss_content = await fetch_hn_rss(feed_url)\n if rss_content.startswith(\"Error\"):\n return rss_content\n\n stories = parse_hn_rss(rss_content)\n\n # Limit to requested count\n stories = stories[: min(count, len(stories))]\n\n if not stories:\n return \"No stories found.\"\n\n formatted_stories = [format_hn_story(story) for story in stories]\n return \"\\n---\\n\".join(formatted_stories)\n```\n\nWith this, here’s the tools that our agent has from both news-mcp and the built-ins:\n\n```markdown\n> /tools\n\nTool Permission\n▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔\nnews_mcp (MCP):\n- news_mcp___get_wired_stories * not trusted\n- news_mcp___get_techcrunch_stories * not trusted\n- news_mcp___get_wallstreetjournal_stories * not trusted\n- news_mcp___get_hackernews_stories * not trusted\n- news_mcp___get_ainews_latest * not trusted\n\nBuilt-in:\n- fs_write * not trusted\n- execute_bash * trust read-only commands\n- report_issue * trusted\n- fs_read * trusted\n- use_aws * trust read-only commands\n```\n\n## Setting up agents to process news\n\nNext, we’ll set up a multi-agent system to parse the news feeds and generate summaries. We’ll have a main agent (image below, top-left) that coordinates three sub-agents, each running in a separate tmux window (image below; bottom-left and right panes).\n\n![The main agent (top-left) with its newly spawned sub-agents](https://eugeneyan.com/assets/news-agents.webp \"The main agent (top-left) with its newly spawned sub-agents\")\n\nThe main agent (top-left) with its newly spawned sub-agents\n\nThe main agent will first divide the news feeds into three groups. Then, it’ll spawn three sub-agents, each assigned to a group of news feeds. In the demo, these sub-agents are displayed as a separate tmux window for each sub-agent.\n\n```markdown\n# Main Agent - Multi-Agent Task Coordinator\n\n## Role\n\nYou are the primary coordinating agent responsible for distributing tasks among \nsub-agents and aggregating their results. \n- Read feeds.txt for the input feeds\n- Read context/sub-agent.md to understand your sub agents\n- Return the summary in the format of context/main-summary-template.md\n\n## Task Assignment Instructions\n\nUse the following message format when assigning tasks: \"You are Agent [NUMBER]. \nRead the instructions at /context/sub-agent.md and execute it. Here are the \nfeeds to process: [FEEDS]\"\n\n...\n```\n\nTruncated instructions for main agent\n\nThen, the sub-agents will process their assigned news feeds and generate summaries for each of them. The sub-agent also categorizes stories within each feed, such as AI/ML, technology, business, etc. Throughout this process, the sub-agents display status updates. When the sub-agent finishes processing its assigned feeds, it displays a final completion status.\n\n```markdown\n# Sub-Agent - Task Processor\n\n## Role\n\nYou are a specialized processing agent designed to execute assigned tasks \nindependently while reporting progress to the main coordinating agent. \nProcess each feed individually and completely before moving to the next one. \nWrite the summaries to summaries/ which has already been created. \nReturn the summary in the format of context/sub-summary-template.md.\n\n...\n```\n\nTruncated instructions for sub-agent\n\nWhile the sub-agents are processing their assigned feeds, the main agent monitors their progress. When all sub-agents are done, the main agent reads the individual feed summaries and combines them into a final summary.\n\n## Defining the news feeds\n\nFinally, we define the news feeds in [feeds.txt](https://github.com/eugeneyan/news-agents/blob/main/feeds.txt) below. We have six feeds: Hacker News, The Wall Street Journal Tech, The Wall Street Journal Markets, TechCrunch, AI News, and Wired.\n\n```markdown\nhackernews: https://news.ycombinator.com/rss\nwsj-tech: https://feeds.content.dowjones.io/public/rss/RSSWSJD\nwsj-markets: https://feeds.content.dowjones.io/public/rss/RSSMarketsMain\ntechcrunch: https://techcrunch.com/feed/\nainews: https://news.smol.ai/rss.xml\nwired: https://www.wired.com/feed/tag/ai/latest/rss\n```\n\nAnd here’s the truncated [main-summary](https://github.com/eugeneyan/news-agents/blob/main/summaries/main-summary.md) it generated for 4th May.\n\n```markdown\n# News for May 4, 2025\n\n### Global Statistics\n- **Total Items Across Sources:** 124\n- **Sources:** 6 (Hacker News, WSJ Tech, WSJ Markets, TechCrunch, AI News, Wired)\n- **Date Range Covered:** May 2-4, 2025\n- **Total Categories Identified:** 42\n\n### Category Distribution\n| Category | Count | Percentage | Top Source |\n|----------|-------|------------|------------|\n| AI/Machine Learning | 31 | 25% | AI News |\n| Business/Finance | 18 | 14.5% | WSJ Markets |\n| Technology | 16 | 12.9% | Hacker News |\n| Politics/Government | 7 | 5.6% | Wired |\n| Cybersecurity/Privacy | 6 | 4.8% | TechCrunch |\n| Trade Policy | 6 | 4.8% | WSJ Markets |\n\n---\n\n## Cross-Source Trends\n\n### Global Top 5 Topics\n\n1. **AI Integration Across Industries**\n - Mentions across sources: 31\n - Key sources: AI News, WSJ Tech, TechCrunch, Wired\n - Representative headlines: \"AI Agents Are Learning How to Collaborate. Companies \n Need to Work With Them\", \"Agent-to-Agent (A2A) Collaboration\", \"Nvidia CEO Says \n All Companies Will Need 'AI Factories'\"\n\n2. **Trade Policy and Tariff Impact**\n - Mentions across sources: 12\n - Key sources: WSJ Markets, TechCrunch, WSJ Tech\n - Representative headlines: \"Temu stops shipping products from China to the U.S.\", \n \"Car Buyers Rushing to Beat Tariffs Find It's Tougher to Get Financing\", \n \"The Future of Gadgets: Fewer Updates, More Subscriptions, Bigger Price Tags\"\n\n3. **Government AI Implementation**\n - Mentions across sources: 7\n - Key sources: Wired, AI News\n - Representative headlines: \"DOGE Is in Its AI Era\", \"DOGE Put a College Student\n in Charge of Using AI to Rewrite Regulations\", \"A DOGE Recruiter Is Staffing a\n Project to Deploy AI Agents Across the US Government\"\n\n4. **AI Safety and Regulation Concerns**\n - Mentions across sources: 9\n - Key sources: TechCrunch, WSJ Tech, Wired\n - Representative headlines: \"One of Google's recent Gemini AI models scores worse on \n safety\", \"AI chatbots are 'juicing engagement' instead of being useful\", \"Dozens\n of YouTube Channels Are Showing AI-Generated Cartoon Gore and Fetish Content\"\n\n...\n```\n\nTry it for yourself! Install [Amazon Q CLI](https://docs.aws.amazon.com/amazonq/latest/qdeveloper-ug/command-line-installing.html) and play with the code here: [news-agents](https://github.com/eugeneyan/news-agents).\n\n```bash\ngit clone https://github.com/eugeneyan/news-agents.git\n\ncd news-agents\nuv sync # Sync dependencies\nuv tree # Check httpx and mcp[cli] are installed\n\nq chat --trust-all-tools # Start Q\n\n/context add --global context/agents.md # Add system context for multi-agents\n\nQ, read context/main-agent.md and spin up sub agents to execute it. # Start main agent\n```\n\nInitially, I wanted to host this as a web app, perhaps on a platform like [Daytona](https://www.daytona.io). However, I quickly learned that building remote MCPs isn’t trivial, especially with only a couple of weekend hours to hack on this. For now, I’ll explore applying this setup to other use cases such as parsing design docs and [COEs](https://aws.amazon.com/blogs/mt/why-you-should-develop-a-correction-of-error-coe/), or multi-agent writing and coding workflows. If you’re also experimenting with MCPs or agentic workflows, I’d love to [hear from you](https://x.com/eugeneyan)!\n\nIf you found this useful, please cite this write-up as:\n\n> Yan, Ziyou. (May 2025). Building News Agents for Daily News Recaps with MCP, Q, and tmux. eugeneyan.com. https://eugeneyan.com/writing/news-agents/.\n\nor\n\n```\n@article{yan2025news-agents,\n title = {Building News Agents for Daily News Recaps with MCP, Q, and tmux},\n author = {Yan, Ziyou},\n journal = {eugeneyan.com},\n year = {2025},\n month = {May},\n url = {https://eugeneyan.com/writing/news-agents/}\n}\n```\n\nShare on:\n\n\n\n\n\n\n\n\n\n\n\nJoin **11,800+** readers getting updates on machine learning, RecSys, LLMs, and engineering.\n", "dateCreated": "2026-09-27T19:29:15.293681927+00:00", "date_created": "2025-05-04T00:00:00+00:00", "date_modified": "2025-05-04T00:00:00+00:00", "date_published": "2025-05-04T00:00:00+00:00", "description": "Learning to automate simple agentic workflows with Amazon Q CLI, Anthropic MCP, and tmux.", "external_url": "https://eugeneyan.com//writing/news-agents/", "git": { "commit": "ebb0941c5c06a2ab2377342d607535262a3e969e", "path": "items/eugeneyan-com/2025/05/2025-05-04-building-news-agents-for-daily-news-recaps-with-mcp-q-and.md", "permalink": "https://github.com/aymericbeaumet/aggr-instance/blob/ebb0941c5c06a2ab2377342d607535262a3e969e/items/eugeneyan-com/2025/05/2025-05-04-building-news-agents-for-daily-news-recaps-with-mcp-q-and.md" }, "id": "urn:aggr:item:3134124baeb646c6b0fac69b95979ec4d384f682", "isBasedOn": { "@id": "https://eugeneyan.com//writing/news-agents/" }, "mainEntity": { "@id": "https://eugeneyan.com//writing/news-agents/", "@type": "CreativeWork", "archivedAt": { "@id": "https://aggr.aymericbeaumet.com/items/eugeneyan-com/2025-05-04-building-news-agents-for-daily-news-recaps-with-mcp-q-and/#webpage" }, "dateModified": "2025-05-04T00:00:00+00:00", "datePublished": "2025-05-04T00:00:00+00:00", "headline": "Building News Agents for Daily News Recaps with MCP, Q, and tmux", "timeRequired": "PT7M", "url": "https://eugeneyan.com//writing/news-agents/", "wordCount": 1530 }, "name": "Archived snapshot: Building News Agents for Daily News Recaps with MCP, Q, and tmux", "preview": null, "reading_minutes": 7, "replicated_at": null, "source": "eugeneyan-com", "source_name": "Eugene Yan", "summary": "Learning to automate simple agentic workflows with Amazon Q CLI, Anthropic MCP, and tmux.", "tags": [ "learning", "llm", "🛠" ], "temporalCoverage": "2026-09-27T19:29:15.293681927+00:00", "title": "Building News Agents for Daily News Recaps with MCP, Q, and tmux", "url": "https://aggr.aymericbeaumet.com/items/eugeneyan-com/2025-05-04-building-news-agents-for-daily-news-recaps-with-mcp-q-and/", "word_count": 1530 }