Google Sheets project summary tool

What it is

This n8n workflow is a tool for an AI agent, and another workflow has to call it. Five nodes run in a line. A Google Sheets node reads every row of a project sheet, and gpt-4o writes a short summary of each project from its name, its notes and its status. Two last nodes gather the summaries into one response field. The caller passes nothing the workflow reads, so the answer always covers the sheet you pointed it at.

Who it suits

Team leads and freelancers who keep their projects in a sheet and want an agent to answer “where do things stand?” without anyone copying rows into a chat. A scheduled flow can call it too and mail the result on.

Where to use it

Any n8n that can reach Google Sheets and OpenAI will do, on n8n Cloud or on a server of your own. Nothing happens until something calls it. An agent’s Call n8n Workflow Tool node is the usual caller, behind a chat or a Telegram bot. An Execute Workflow node in a weekly schedule works as well.

Setup

Get it saves a zip with workflow.json and Mindket’s setup notes in your downloads folder. Have two things ready: an OpenAI key, and a Google sheet with one row per project and columns for the name, the notes and the status.

1. In n8n, create a workflow, open its menu, choose Import from File and pick workflow.json.
2. Add an OpenAI credential and select it in the OpenAI node.
3. Add a Google Sheets OAuth2 credential and select it in Google Sheets.
4. In the same node, choose your spreadsheet in Document, which holds YOUR_SHEET_ID, then the tab in Sheet.
5. Make the column names match. The OpenAI message reads Project Name with a trailing space, then Notes and Status, as its author’s sheet had them. Rename your headers or edit the message.
6. Save, then add the node that runs it to the calling workflow. For an agent, give the tool one line saying it returns a summary of every project in the sheet, because the agent reads that line to decide when to use it.

The caller passes nothing the workflow reads, and one text field comes back as response:

in:  { }
out: { "response": "<a summary of each project>" }

How to use it

Ask the agent how the projects are doing. It calls the tool, the tool reads the sheet fresh each time, and the agent answers from response. To try it without an agent, run the workflow from the editor with nothing pinned on Execute Workflow Trigger.

Example

A sheet holds two projects.

Each row goes to OpenAI on its own, so the model writes two summaries, one for each project. Aggregate gathers both into response, and the agent turns them into its answer.

For newcomers and experts

If you are new. A sub-workflow is an ordinary workflow that starts from Execute Workflow Trigger and waits to be called. Run it on its own first, with a small sheet, and read what each node sent on.

For experts. Google Sheets has no filter, so each row costs a gpt-4o call, and a long sheet makes a long reply for the agent to read. Add a filter on Status to leave finished projects out. The message in OpenAI carries only three columns, so add any others to it by name.

Good to know

Every row you leave in the sheet goes to OpenAI, so keep private columns out of the message.

• Billing: one OpenAI call per row
• Read only: the workflow reads the sheet and writes nothing back
• Header names: the expression expects Project Name with a trailing space, so a sheet headed Project Name gives the model no name until you edit it
• Credentials: the JSON holds none, and nothing from its author’s account