Fort Smith Newsletter Automation
End-to-end newsletter automation using n8n, Claude, and custom MCP servers. Curates local news and events for Fort Smith, Arkansas with 130+ hours saved annually.
TL;DR: TL;DR: I built an end-to-end newsletter automation system for a Fort Smith, Arkansas newsletter. Uses n8n for orchestration, Claude for content generation, and custom MCP servers for news/events curation. Saves 130+ hours annually.
The Problem
Running a local newsletter is time-consuming:
- News research: 2-3 hours finding relevant local stories
- Events curation: 1-2 hours gathering upcoming events across sources
- Content writing: 2-3 hours drafting, formatting, editing
- Quality control: 1 hour reviewing before send
That's 6-9 hours per newsletter. With weekly sends, that's 300-450 hours/year just on production.
My Approach
I built an automation pipeline with three main components:
- News Aggregation: Custom MCP servers that search local news sources by zipcode
- Events Curation: Automated scraping and categorization of local events
- Content Generation: Claude-powered drafting with consistent voice and format
The system runs on n8n, which orchestrates the workflow and handles scheduling.
Architecture
Fort Smith Newsletter Automation - Architecture Diagram
Key Features
- Zipcode-Based News: Targets Fort Smith, AR (72901) and surrounding areas
- Event Categorization: Music, Family, Business, Fitness, Arts, Outdoor
- Direct Event Links: Every listing includes source URL
- Weather Integration: Local forecast included in each issue
- Consistent Format: Featured stories + Quick updates + Events sections
- Human Review Step: Draft generated, human approves before send
Results & Metrics
| Metric | Before | After | Improvement |
|---|---|---|---|
| Production Time | 6-9 hours | 1-2 hours | ~75% reduction |
| News Sources Checked | 3-5 | 15+ | 3x coverage |
| Events Found | 10-15 | 30-50 | 3x more |
| Annual Hours Saved | - | 130+ | - |
| Cost per Issue | - | ~$2-3 (API) | - |
What I Learned
The biggest challenge was maintaining consistent voice. Early versions sounded robotic or inconsistent between sections. I solved this by:
- Detailed system prompts with specific style guidelines
- Few-shot examples from previous newsletters
- Section-specific instructions (news is informative, events are promotional)
Another insight: human review is essential. The automation handles 90% of the work, but a human editor catches factual errors, updates last-minute changes, and adds personal touches that AI misses.
The MCP servers I built for this project became reusable tools:
google-news-mcpfor local newsfort-smith-events-curatorfor eventsfort-smith-news-generatorfor content
Frequently Asked Questions
What problem does this automation solve?
It reduces newsletter production from 6-9 hours to 1-2 hours per issue. The automation handles news research, events curation, and initial drafting, leaving humans to review and approve.
What technologies power this project?
n8n for workflow orchestration, Claude for content generation, custom MCP servers for news/events collection, and Perplexity API for supplementary research.
How accurate is the automated content?
Highly accurate for facts (news links, event details, dates). The system always includes source links for verification. A human review step catches any errors before publishing.
Frequently Asked Questions
More Projects
View allBuilt by Abhinav Sinha
AI-First Product Manager who builds production-grade tools. Passionate about turning complex problems into elegant solutions using AI, automation, and modern web technologies.