Hey everyone 
I’m currently building a modular AI assistant system that connects voice input (via Telegram), task and project data (Notion), calendar events (Google Calendar), and email workflows (Gmail).
My goal is to create an intelligent, voice-controllable agent for everyday use — especially for clients who want to quickly create tasks, notes, calendar blocks, or email drafts just by speaking.
Here’s the structure I’m currently using (see attached mindmap for visual):
Architecture overview:
- Main Agent: routes incoming requests to specialized agents
- Notion Agent
- Create Tasks Agent
- Create Meetings Agent
- Create Thoughts / Notes Agent
- Research Agent (searches Notion inside databases)
- Search Project & Customer DB
- Search Meeting DB
- Search Task DB
- Calendar Agent: creates calendar entries based on extracted date/time from voice
- E-Mail Agent: creates Gmail drafts based on voice input
What I’m trying to figure out:
- What is the best way to manage and orchestrate these agents in n8n?
- Should I use a single AI agent with a routing function (intent detection), or keep each sub-agent separate with dedicated nodes and prompts?
- How can I make this more resilient, especially when e.g. no project or customer is identified?
- Any ideas how to preserve long-term memory per client, ideally with some lightweight memory (e.g. JSON or Notion thread)?
Potential issues I see:
- Managing fallbacks (e.g. “What project is this for?” if not found)
- Avoiding prompt duplication across agents
- Handling ambiguous input like: “Send a follow-up to Customer A, create task for X and create meeting for date Z” – can it handle all requests at once?
If you’ve built something similar or if you’ve found a clean way to chain and scale AI agents with context in n8n - I’d love to learn from your setup or suggestions 
Thanks in advance!
