Two small Claude-powered agents that automate the top of the content operations funnel for a niche newsletter: trending-topic research and YouTube content strategy. Built to seed The Emirates Allocator newsletter — UAE crypto — but the pattern generalises to any niche.
Runs a structured research brief against Claude and writes a dated markdown report.
- Top 5 trending topics in UAE crypto
- New pain points surfacing for UAE residents
- Regulatory updates (VARA, UAE Central Bank)
- Content recommendations scored by search demand × monetization × content gap
Same pattern, tuned for the YouTube surface.
- Top-performing topics and formats in UAE crypto
- Content gaps a new creator can own
- Channels to study
- 10 specific video ideas with hook title + thumbnail concept + talking points
- CPM and sponsorship monetization notes
export ANTHROPIC_API_KEY=sk-ant-...
pip install -r requirements.txt
python research_agent.py
python youtube_research_agent.pyReports land in ./output/pain-points/ and ./output/youtube/ as dated .md files.
- Structured prompt, free-form output. Prompts enforce a report shape; the model fills it. Cheaper and faster than chaining multiple calls, and the output is human-reviewable in seconds.
- Env-var config.
ANTHROPIC_API_KEY,OUTPUT_DIR,ANTHROPIC_MODEL— no secrets in code, no hardcoded paths. - Dated files. Every run is immutable — easy to diff week-on-week for trend drift.
- Said no to. RAG, web-scraping, embeddings, vector DBs — a single prompt + fresh model call is enough for weekly cadence content research. Save the complexity budget for downstream production.
Content research is the slowest, most expensive part of running a niche newsletter when done by hand. These agents turn a Monday morning's research into a 2-minute cron job — the editorial team spends the time on judgment and voice, not link-hunting.
MIT