Before
They had already built a personal AI agent team in Make.com, which they talked to only through Slack. But autonomous agents were a fast-moving, widely discussed topic, and they wanted real experience with other platforms. Flowise, met during the course, was a start, but they wanted to understand how agent teams are built across tools.
“So far I liked n8n the most, so the goal was to use n8n to create an AI agent team with a supervisor and multiple subagents”
In their own words
The turning point
The AI Mastery course opened the door with Flowise, and from there they tested several other tools. n8n stood out. They set a clear goal: a supervisor agent with subagents handling tasks that mattered in their own day, learning along the way with ChatGPT as a sparring partner and YouTube for research.
“By experimenting, using ChatGPT as my sparring partner, and conducting research on YouTube, I was able to build an AI agent team capable of handling several tasks that are relevant to me.”
In their own words
After
They built a working team in n8n. A supervisor agent uses an Airtable contact list and directs an Email Agent, a Calendar Agent, an Expense Tracker Agent, a To-Do List Agent and a Perplexity research tool. Memory keeps the context of earlier messages, and Telegram accepts both text and audio. Early tests delivered reliable results, and they plan to keep expanding it.
What it lets people do
One message can ask for an overview of the last two emails from a contact, book a calendar event with participants, log an expense or add a to-do. The agent looks up the right email addresses on its own, so routine admin becomes a short chat.
What they learned
They learned how tools, memory and chat models fit together inside n8n. Just as much, they found that experimenting, asking ChatGPT and searching YouTube can carry a project a long way when time is short. Now they look forward to comparing n8n and Make.com for different uses.
“It is great to see what is possible to achieve and to set up.”
In their own words
Autonomous AI Agent Team in n8n was built during AI Mastery.
Read the full submission
Overview
In this project, the goal was to explore tools for creating autonomous AI agents and agent teams and to pick one for creation of an autonomous AI Agent team.
Besides Flowise, which we encountered during the AI Mastery course, I explored several other tools and experimented with them. So far I liked n8n the most, so the goal was to use n8n to create an AI agent team with a supervisor and multiple subagents that could perform the following tasks:
- Conduct research via Perplexity
- Create entries in my Airtable To-Do list
- Add entries to my Airtable expense tracking database
- Send emails
- Provide an overview of emails received from specific individuals (e.g., "Give me an overview of the last two emails from Peter Bauer")
- Retrieve an overview of calendar events scheduled for a specific time period
- Create new calendar entries and events (including participants)
The foundation for calendar and email activities is an Airtable database containing my contacts (sample data has been created for the demo). This database is accessed to identify the relevant email addresses for the respective persons mentioned in the request. As communication channel Telegram has been chosen as I do not have a WhatsApp Business Account.
Agents or Tools Assigned to the "Supervisor Agent"
- Contact List in Airtable
- Email Agent for email-specific tasks
- Calendar Agent for calendar-specific tasks
- Expense Tracker Agent
- To-Do List Agent
- Perplexity Research Tool (includes HTTP request)
- Initially, I used SerpAPI for research tasks but later replaced it with Perplexity.
- SerpAPI (Research AI Agent has been activated for now).
Memory and Communication:
- A Windows Buffer Memory was implemented to ensure it has context and memory from previous messages
- OpenAI Chat Model was selected as the chat engine.
- Telegram was chosen as the communication channel.
- Both text and audio inputs are supported via Telegram
This setup enabled the supervisor agent to orchestrate the subagents, ensuring a cohesive workflow for the specified tasks.
Link to loom video for a quick demo
https://www.loom.com/edit/765354f7df8f47e8b9588fcad63ff080
Sample data has been used to show the workflow.
Learnings and Conclusion
For me, it was important to gain experience with tools like Flowise, n8n, or similar platforms, as well as to develop an understanding of how to build autonomous AI agents or AI agent teams using these tools, as this has become a widely discussed and trending topic.
Through this project, I learned how to work with n8n and gained insight into integrating tools, memory, and chat models. By experimenting, using ChatGPT as my sparring partner, and conducting research on YouTube, I was able to build an AI agent team capable of handling several tasks that are relevant to me. Telegram serves as the communication channel.
The agent team still will be expanded and customized further with additional subagents and tasks to perform. As time was limited, I was not able to complete it.
Since I’m a huge fan of Make.com and have also built my personal AI agent team there, which I communicate with exclusively via Slack, I find it exciting to make a comparison between the two platforms and the different use cases.
The n8n agent team was just completed today (as it is for now), but will be expanded with additional tasks and functionalities within the next days. It hasn’t been used extensively so far. However, the initial tests delivered reliable and good results.
I'm excited to create, learn and implement more in this direction! It is great to see what is possible to achieve and to set up.
Thank you to the AI Mastery Program team and the coaches for showing us the potential of AI Agents and AI Agent teams!!!
Written 21 January 2025· Submitted by the storyteller
Every story is written by a Mindvalley student or by someone our teachers have worked with. How we collect and check stories
