Skip to content

How to INSTANTLY Create n8n Agents Using ChatGPT (No Code)



Gumroad Link to Assets in the Video: https://bit.ly/4gSHnZu Apply to join the Early AI-dopters Community: https://bit.ly/3ZMWJIb …

Tags:

30 thoughts on “How to INSTANTLY Create n8n Agents Using ChatGPT (No Code)”

  15. @janalgos

    would be cool to see a similar approach but for writing cursor rule files. Theres a website that has a bunch of user provided cursor rules, similar to the website with user provided n8n templates…so perhaps could adapt the approach for create epic cursor rule files

  22. @lakegeorge11

    Hi Mark, I have a question:

    The example you did with creating Jarvis via ChatGPT Deep Research, was already proven it could be made in N8n (as other people have built this automation before), how would your workflow and prompts change if you were trying creating an automation that no one has built before? / you dont know if what you want to automate is even possible to be made in an automation? … how would your process change for this?

  27. @MomVersusTheWorld

    Here is 1St prompt for anyone who wants it. Question: Would it make sense to create GPT or SPACE in Perplexity to easily access this prompt? If not, why?

    Do full research on how to build n8n workflows using JSON. I want to be able to use an LLM to create an entire automation flow in n8n.

    Research the newest builds and module builds that will let you create all kinds of triggers and workflows, just through writing JSON formatted perfectly to be imported in n8n.

    Ensure the following in your output JSON:

    1) PropertyName issues: There are no reasons to get propertyName is property value errors, and you generate perfect syntax.

    2) Comments Removed: No inline comments (using //) cause the JSON to be invalid.

    3) Expressions: The expressions using ={{ … }} remain intact. These are valid in n8n’s JSON configuration as they are processed at runtime.

    4) Structure: The structure with "nodes" and "connections" is preserved as expected by n8n.

    Here's the exact workflow I'd like to create in n8n:

    Trigger → Telegram Trigger (Receives message)
    Decision → Switch (Checks if message is audio, text, or error)

    If Audio → Get Voice File
    → Speech to Text (Transcribe Recording)

    If Text → Edit Fields (Manual processing)

    Processing → J.A.R.V.I.S (AI Tools Agent)

    Uses:
    – OpenAI Chat Model
    – Window Buffer Memory
    – Gmail (get message)
    – Get Calendar (get all events)
    – Contacts (search record)
    – Tasks (search record)
    – Create Contact (create record)
    – SerpAPI
    – Hacker News (get all)
    – Calculator

    AI Model → Basic LLM Chain
    Response Handling

    Text Response → Send Telegram Message (Text)
    Audio Response →
    – Text to Speech (API call to ElevenLabs)
    – Send Telegram Message (Audio)

    Alternative Processing → Uses Anthropic Chat Model

  28. @cj708

    Great video. Thanks for this. Some questions…Would it make sense to upload some crucial n8n documentation to the chatgpt project vs. all templates? Are you able to pull the list of nodes from the current n8n github version using lovable or bolt then feed it to chatgpt so it will have an inventory of all of the nodes available? I think the ultimate solution is have your n8n co-pilot create a n8n workflow that generates n8n workflows. Use the vector component of supabase as a knowledge base, then have the AI Agent access the knowledge base (like vector shift).

Comments are closed.