AI Newsletter System — Multi-Agent Research & Writing Workflow
A scheduled n8n workflow where planner, writer and editor AI agents research with Tavily, write cited sections and save a polished HTML newsletter as a Gmail draft.
Built by Redoan Uzzaman, Full-Stack Developer & AI Engineer.

Overview
The AI Newsletter System automates the research and writing behind a recurring newsletter. On a schedule, it gathers recent news for a chosen topic, plans the issue, writes it and leaves a ready-to-review draft in Gmail.
How it works
The workflow is organized as a small team of AI agents connected through OpenRouter. A schedule trigger starts an initial Tavily search on the main topic. A planning agent (GPT-5 Mini) reviews the results and produces a newsletter title plus three focused sub-topics, returned as structured JSON validated by an output parser.
The sub-topics are split into separate items, and each one gets its own Tavily deep-dive followed by a section-writer agent that drafts that section with source citations. The three sections are then aggregated and passed to an editor agent (GPT-5), which unifies tone, structure and flow, consolidates and deduplicates references into a single alphabetized list, keeps the issue under 1,000 words and formats it as clean, email-ready HTML.
The final step creates a draft in Gmail rather than sending automatically, keeping a human in the loop for review before publication. Topics, schedule and word limits are easy to adjust for any niche.
Key Features
- Scheduled weekly or monthly runs
- Tavily-powered web research per topic
- Planner agent producing title and three sub-topics
- Parallel section-writer agents with citations
- Editor agent for cohesive, polished HTML output
- Deduplicated, alphabetized source list
- Gmail draft creation for human review
Tech Stack
- n8n
- OpenRouter
- OpenAI GPT-5
- Tavily
- Gmail API
- LangChain
Want something like this built?
Redoan Uzzaman takes on AI, automation and full-stack projects for clients worldwide.