Chat copywriter for ads, ad scripts and cold emails

What it is

This n8n workflow is a chat that writes marketing copy and files it as a Google Doc. You ask n8n’s chat for ad copy, a video ad script or a cold email sequence. An agent on gpt-4.1-mini picks one of three tools, which call writing steps further down the same workflow. The finished copy goes to a Markdown-to-Google-Docs converter that you supply. The file came with eighteen nodes. Mindket rebuilt three and added one, because the OpenAI service they called has closed.

Who it suits

Marketers and small agencies who write many short pieces and want each one saved as a document they can share. The last step needs a converter of your own. Skip it if you have none and will not build one. Without it, the copy stays in the run’s data.

You need some n8n: importing a file, two credentials, three tool links and one converter address. Its author puts setup at about 10 minutes.

Where to use it

Any n8n with its chat open to you or your team. The doc lands in the Google Drive of the account behind the Google Docs credential.

• n8n Cloud: import the file into a new workflow there
• Your own n8n server: the same file, with your converter on a server you control
• n8n’s chat: the canvas chat while you test, or a public chat page once you turn that on in the trigger
• Google Docs: one document for each request, titled by the agent

Setup

Get it saves a zip with workflow.json and Mindket’s setup notes in your downloads folder. Set up the converter first, since the last node needs its address.

1. In n8n, create a workflow, open its menu, choose Import from File and pick workflow.json.
2. Save it, so the workflow has an ID of its own.
3. In Ad Copy Agent, Cold Email Copywriter and Ad_script_agent, pick this same workflow from the list. Each tool calls the lower half of this file.
4. Add an OpenAI credential and select it in OpenAI Chat Model and OpenAI Chat Model1.
5. Add a Google Docs OAuth2 credential and select it in HTTP Request.
6. In HTTP Request, replace YOUR_CONVERTER_URL with your converter’s address.
7. Save, open the chat on the canvas and ask for one piece of copy.

The original posted the copy, with your Google Docs token, to a converter on its author’s own server. That address is gone. Yours receives a JSON body with fileName and output, the Markdown, and your Google token in a Bearer header. Run it somewhere you control, or swap HTTP Request for n8n’s Google Docs node. That node creates a document and inserts text, with the Markdown marks shown as typed.

Its three writing nodes called assistants kept in the author’s OpenAI account. OpenAI’s deprecations page, read on 2 November 2026, gives 26 August 2026 as the shutdown date of the Assistants API. So each is now an AI Agent of the same name with a short system message of Mindket’s, and the three share the new OpenAI Chat Model1.

How to use it

Type a request in the chat, with the product, the audience, the offer and the tone. The agent passes it to one tool with a label: ad_copy, ad_script or cold_email. The writer answers, the agent wraps the copy in JSON with a title, and Code1 reads it. Then the converter makes the doc. The chat’s reply is whatever your converter answers, so have it return the document’s link.

Example

You type this in the chat.

Three Facebook ads for a yoga studio. Offer: first month half price. Audience: office workers. Tone: calm.

The agent calls Ad Copy Agent with ad_copy and your request, and Switch sends it on to Ad copy agent, which writes the three ads in Markdown. The agent then answers in this shape, and Code1 parses it:

{
  "toolName": "ad_copy",
  "response": {
    "title": "…",
    "markdown": "### Ad 1\n…"
  }
}

HTTP Request sends the title as fileName and the Markdown as output to your converter, which creates the doc.

For newcomers and experts

If you are new. n8n draws a workflow as boxes joined by lines, and each box, a node, does one job. This file holds two workflows in one. The chat sits on the top line, and the writing steps its tools call sit on the bottom line. Ask for one short ad first. Then open both runs under Executions, the chat’s and the tool’s, to see what each half did.

For experts. The router’s three tools call this same file, and Switch picks the writer by the toolName the model sends.

A label the switch does not know ends the tool run with nothing. Code1 fails when the router adds words around its JSON.

Good to know

It needs your own converter, and the router has to answer in clean JSON.

• Billing: OpenAI bills the router and one writer for each request, and the router may call a tool more than once
• Your Google token: the converter gets a token that can edit your Google Docs. Run it yourself, or swap in n8n’s Google Docs node
• Strict JSON: Code1 parses the router’s whole answer, so a sentence before the JSON stops the run
• The writers’ prompts: the original’s lived in its author’s OpenAI account and are not in the file. The system messages here are short starting points, so add your brand’s voice
• Model date: OpenAI’s deprecations page lists GPT-4.1 mini (gpt-4.1-mini-2025-04-14) for removal from the API on 14 January 2027, with gpt-6-luna as its replacement
• Your data: every request goes to OpenAI, and the finished copy goes to your converter
• Credentials: the JSON holds none. The three tools point at no workflow until you pick this one