This is a submission for the Notion MCP Challenge
What I Built
notion-newsroom — An AI-powered newsroom automation system that enriches breaking news articles with historical context in real-time.
When journalists publish a news article in Notion, the system automatically:
- Generates semantic search queries using Gemini LLM to understand the article's core topics
- Retrieves relevant historical context from a vector database of past articles, archives, and source materials using ChromaDB
- Appends enriched context blocks directy to the Notion page with citations, relevance scores, and source links
- Provides explicit feedback via Notion comments, so journalists know what context was found or why it wasn't available
This dramatically accelerates research workflows — journalists no longer need to manually search archives or context databases. The system runs on a schedule and handles mixed data sources (Notion articles, CSV archives, RSS feeds) seamlessly.
Core features:
- Automated context discovery with configurable relevance thresholds
- Rich Notion block composition with collapsed toggle blocks for clean UI
- Explicit error reporting (query generation failures, search unavailable, no relevant matches)
- Support for heterogeneous archive sources (UUIDs, short numeric IDs)
- Transactional append safety with built-in chunk size validation
Video Demo
Show us the code
GitHub: udofia2/notion-newsroom
How I Used Notion MCP
Notion MCP is the core integration that makes this workflow possible. Here's how:
1. Reading Published Articles
The MCP server exposes Notion page content as queryable objects. The Context Hunter reads drafts and published articles from a Notion database, extracts content, and pipes it into the semantic search pipeline.
# Read article from Notion via MCP
article_content = notion_client.retrieve_block(page_id)
2. Appending Historical Context Blocks
Once context is found and ranked, MCP enables real-time append of rich text blocks to the Notion page:
- Composition of nested blocks (toggle → bulleted items → paragraphs with citations)
- Support for Notion mentions (linking to source pages) and plain text links (for external archives)
- Automatic chunking to respect Notion's per-element constraints
# Append context block with MCP
context_block = build_historical_context_toggle_block(results, article_title)
notion_client.append_block_children(page_id, [context_block])
3. Posting Automation Feedback
Journalists get explicit comments on articles for all outcomes:
- ✅ "Historical context added: 8 items"
- ⚠️ "Query generation failed: provider returned server error 503"
- 🔍 "Search unavailable: embedding model offline"
- ❌ "No relevant historical context found"
This visibility is critical — silent automation failures destroy trust. MCP comments bridge the gap.
4. Notion Mentions & Links
The system intelligently handles mixed source IDs:
-
Notion page UUIDs → Rich mentions (
mention.page.id) - External/Archive sources → Text links with URLs
- Short numeric IDs → Plain text citations with metadata
