Daily trend tracker from X accounts and websites

Daily trend tracker from X accounts and websites

An n8n workflow that runs at 6 every morning on a topic you name. It reads the last day of posts from X accounts you list
and the news on websites you list, then saves the items worth knowing to a Google Sheet. Apify fetches the posts, an
Airtop browser reads the sites, and two OpenAI agents read and choose.

What is inside

• workflow.json: the workflow as n8n exports it, 26 nodes. Three are sticky notes that label its parts: the X posts, the
  websites, and the analysis with the save. It holds no credential.
• README.md: these notes.

What we changed from the original copy

• The author's spreadsheets are gone. Get X Accounts pointed at one copy of the sheet, and Get Websites and Save at
  another. All three now ask for the same YOUR_SHEET_ID and find their tab by name: X, Websites or Data Points.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• An OpenAI API key that can use gpt-5. Read Good to know about its shutdown date.
• An Airtop account and its API key.
• An Apify account and its API token.
• A Google account, and a Google Sheet with three tabs (below).

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Make the sheet with the three tabs below, then copy its ID: the long part of its address between /d/ and /edit.
3. Add a Google Sheets OAuth2 credential. In Get X Accounts, Get Websites and Save, select it and replace
   YOUR_SHEET_ID with the ID.
4. Add an OpenAI credential and select it in OpenAI and GPT.
5. Add an Airtop credential with your API key and select it in Session, Window, Load URL, Query and
   Terminate a session.
6. Add a Header Auth credential for Apify. Its name is Authorization, and its value is Bearer, a space and your token.
   Select it in Get Tweets.
7. Open Your Instructions and replace [ Your Topic ] with your subject, once in Browser Prompt and three times in
   Analysis Prompt.
8. Put a few handles in X and a few page addresses in Websites.
9. Save, press Execute workflow once to test it, then press Publish so it runs each morning.

The sheet

| Tab | Columns |
|---|---|
| X | username: one X handle for each row, written as it appears after the @ |
| Websites | website: one page address for each row, such as a newsroom or a blog's front page |
| Data Points | name, description, source, date. The workflow adds the rows |

How a run goes

1. Every Day 6 AM starts the run at 6 in the morning, in the workflow's time zone. Your Instructions holds the two
   prompts the agents read later.
2. Get X Accounts reads the X tab. For each handle, Get Tweets asks the Apify actor CJdippxWmn9uRfooo for the 20
   latest posts in English.
3. Select Fields keeps four fields: username, text, createdAt and url. Filter drops posts older than 25 hours,
   and Aggregate Tweets gathers the rest into one list.
4. Session starts an Airtop browser session with the profile name general. Window opens a window on OpenAI's news
   page, with a live view you can watch.
5. Get Websites reads the Websites tab, and Browser Agent takes one site at a time. Through Load URL and Query it
   opens the page and its news pages and keeps what came out in the last 24 hours. It lists each item with a title, a date,
   a description and its address, and JSON1 holds its reply to that shape.
6. Aggregate Websites gathers every site's list, and Terminate a session closes the browser.
7. AI Agent reads both lists, with Analysis Prompt as its instructions, and returns the items that pass, each with a
   name, a description and a source. JSON holds its reply to that shape.
8. Split Out turns the list into one item for each entry, and Save adds each one as a row to Data Points.

Change the look

• What counts as news: Analysis Prompt in Your Instructions. Its include and exclude lists decide what reaches the
  sheet, and it caps a name at 50 characters and a description at 1,000.
• What the browser looks for: Browser Prompt, plus the date line and the output format in the text of Browser Agent.
• Which posts: maxItems, lang and the filter flags in the JSON body of Get Tweets, and the 25 hours in Filter.
• When it runs: Every Day 6 AM.
• Models: OpenAI for the browser agent and GPT for the analysis, both gpt-5.

Good to know

• Model shutdown: OpenAI's deprecations page, read on 2 November 2026, lists gpt-5 for shutdown on 11 December 2026,
  with gpt-5.6-sol as its replacement. Change both model nodes before then.
• A quiet day ends early. If no listed account posted inside the window of Filter, it passes nothing and the run stops
  there, before any website is read. To read the sites anyway, turn on Always Output Data in the settings of Filter.
• The date column stays empty. Save has it, but AI Agent returns only a name, a description and a source. Add a
  date field to the example in JSON and to the prompt's output format to fill it.
• A missing tool: Analysis Prompt tells the agent it can search the web. AI Agent has no tool, so it works from the
  posts and pages alone.
• Shared memory: Simple Memory keeps the browser agent's last ten exchanges under one fixed key, 239. Every site and
  every run share it while n8n keeps running, so notes from one site can carry into the next. Set the key to the site's
  address if that matters to you.
• An open browser: Terminate a session runs only after the last site. When Browser Agent fails, the Airtop session
  stays open until its idle timeout, which the node sets to 10 minutes.
• Costs: OpenAI bills the analysis once a run and the browser agent for every step it takes on every site. Airtop bills
  the session and each query. The Apify actor is kaitoeasyapi's Tweet Scraper. Its free tier listed $0.25 for 1,000 posts
  when we read it on 2 November 2026, and each handle asks for as many as maxItems allows.
• Where the data goes: the posts and the page text go to OpenAI, and the chosen items land in your sheet.