LinkedIn post drafts from a Google Sheets topic queue

LinkedIn post drafts from a Google Sheets topic queue

An n8n workflow. Keep a list of topics in a Google Sheet. Each run takes the next one marked To Do, searches the web
for it with Tavily, has an AI agent on OpenRouter write a short LinkedIn post, and puts the draft back in the row.
Nothing is posted to LinkedIn.

What is inside

• workflow.json: the workflow as n8n exports it, 7 nodes, one of them a sticky note. It holds no credential. The
  spreadsheet and the Tavily key are left as placeholders.
• README.md: these notes.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• A Tavily API key.
• An OpenRouter API key.
• A Google account, and a spreadsheet with a tab named Sheet1 and these column names in row 1: Topic, Status and
  Content.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Fill the sheet with one topic per row in Topic and To Do in Status. Leave Content empty. Keep each topic
   unique, because the draft finds its row again by the topic.
3. Copy the spreadsheet's ID, the long part of its address between /d/ and /edit. Add a Google Sheets OAuth2
   credential, select it in Get Topic and Send Content, and replace YOUR_SHEET_ID with the ID in both.
4. In Tavily, replace <YOUR_TAVILY_API_KEY> in the Authorization header with your key. The brackets go too.
5. Add an OpenRouter credential and select it in OpenRouter Chat Model. Then pick a current model in the same node.
   The file names anthropic/claude-3.5-sonnet, which was not in OpenRouter's model list on 2 November 2026.
6. Save and press Execute workflow. The trigger node is named When clicking ‘Test workflow’.

How a run goes

• Get Topic returns the first row whose Status is To Do.
• Tavily sends the topic with the words "Search the web for" in front of it. It asks for a basic search, three
  results and a short answer, which nothing reads.
• Content Creator is an AI agent with no tools. It gets the content of the three results as Article 1, 2 and 3. Its
  system message asks for an upbeat post for entrepreneurs, at most 700 characters long, with 2 to 4 hashtags and 1 to
  3 emojis. It must not quote the articles or name their sources.
• Send Content finds the row by its Topic and writes the post in Content and Created in Status.

Change the look

• Audience and tone: the system message in Content Creator, written for entrepreneurs.
• Length, hashtags and emojis: the same message.
• Search: search_depth, max_results and topic in the body of Tavily. The agent reads exactly three results, so
  keep max_results at three or more.
• Recent results only: time_range in the same body (below).
• Status words: the To Do filter in Get Topic and the Created value in Send Content.

Good to know

• Each run drafts one topic. Press Execute workflow once per topic, or swap in a Schedule Trigger for the manual one
  to draft a post every morning. A run that finds no row marked To Do stops after Get Topic.
• Nothing reaches LinkedIn. Read each draft in the sheet, make it yours and post it, or add a LinkedIn node after
  Send Content.
• The search is not held to recent pages. The body sends "days": 7 and "time_range": null, and Tavily's search
  reference, read on 2 November 2026, lists time_range (day, week, month or year) and no days field. Set
  "time_range": "week" to keep results to the last week.
• If Tavily returns fewer than three results, the prompt asks for an article that is not there. That article comes
  through empty, or the agent stops with an error, depending on your n8n version.
• Every draft aims at entrepreneurs in an upbeat tone, whatever the topic, so change the system message if your
  readers are someone else.
• Tavily and OpenRouter each bill their own part of a run.
• The Tavily key sits in the request header as plain text, so it travels with any export of the workflow. Take it out
  before you share the file.