App Logo
Productivity

Google Sheets MCP for AI agents: read and edit your sheets from Claude or ChatGPT, no Google Cloud project (2026)

Prompt template

Add these leads to my Google Sheet sheet, then mark the ones I already contacted as Sent.

Get started for free

Overview

A spreadsheet is where a lot of work ends up: the lead list, the tracker, the report someone checks every Monday. An AI agent can find the leads and draft the report, but getting the result into your Google Sheet usually means someone copies it in by hand.

Why it's hard

Every usual route to a Google Sheets MCP starts in Google Cloud. Google's own Sheets MCP server is in Developer Preview and needs a Cloud project, two APIs, an OAuth consent screen and a client ID. Open-source servers ask for the same OAuth credentials or a service account. Hosted connector platforms skip that setup by holding the OAuth tokens for you, so your access to the sheet runs through their service. Apps Script web apps work, but you write and deploy the script yourself.

How Reduck does it

Reduck is an MCP server whose Google Sheets scripts run in your own Chrome, where you are already signed in to Google. Your agent works on exactly the sheets you can open, as you, and sees the same thing you see. There is no Cloud project, no OAuth app and no token held by anyone else. A write answers only once Google has saved it, and when the sheet refuses a value the run stops with the sheet's own message.

What your agent does

  • Read a whole tab as rows and columns
  • Write a block of values from any cell, with formulas and number parsing like typing
  • Clear a range, keeping the formatting
  • List the tabs of a spreadsheet, add, rename or delete one
  • Create a new spreadsheet, rename it, or export it as xlsx, csv, pdf and more

Ways to give an AI agent your Google Sheets

Google's Sheets MCP Hosted connector (Composio, StackOne) Apps Script web app Reduck
Google Cloud project Yes No No No
OAuth app or service account Yes Managed for you No No
Who holds the access You The platform Your script You, in your browser
Writes cells Yes Yes What you code Yes
Keeps your dropdown rules Not documented Not documented What you code Yes, refused values stop the run
Status Developer Preview Generally available Your own code Generally available

Who it's for

  • Founders and sales teams who keep leads and pipelines in Google Sheets
  • Operations and finance people who update a tracker or a report every week
  • Developers who want an agent to write to a sheet without a Cloud project

FAQ

Yes. Google offers a remote Sheets MCP server in Developer Preview. It needs a Google Cloud project, the Sheets and Sheets MCP APIs turned on, an OAuth consent screen and a client ID. Reduck needs none of that, because its Sheets scripts run in your own signed-in Chrome.
No. There is nothing to create in Google Cloud and no key to store. You stay signed in to Google in Chrome, and the Reduck extension runs each script there, as you.
Both. It reads a tab, writes a block of cells, clears a range, adds, renames and deletes tabs, creates a spreadsheet and exports one. Reading changes nothing; every script that changes the sheet is marked as a write, so your agent can ask you first.
Yes. A value is entered the way you would type it, so the sheet's own rules apply. A value a dropdown does not allow is refused, the run stops and reports the sheet's own message, and the dropdown stays in place.
Like the Sheets API's values.update with user-entered input. A missing value leaves the cell as it is, an empty one clears it, and anything else is parsed as if typed, so "=SUM(A1:A3)" becomes a formula and a number becomes a number. Start a value with an apostrophe to keep it as text.
Yes. Reduck connects to ChatGPT, Claude, Claude Code, Codex, Cursor and any other MCP client, and to a terminal through the Reduck CLI. You ask in plain words and the agent picks the Sheets scripts it needs.
Claude's own Google Workspace connector edits Sheets inside Claude. Reduck's scripts are the same wherever you call them from, ChatGPT, Codex, Cursor, a CLI or your own agent, and they combine with scripts for other sites, such as reading leads from LinkedIn and writing them to a sheet in one run.
No. The scripts use the Sheets editor like a person would, so a block of a few rows takes several seconds rather than milliseconds. For bulk jobs of thousands of rows a service account with the Sheets API is faster; for an agent that updates a sheet as part of its work, this is set up in a minute.

Similar use cases

Google LinkedIn

LinkedIn lead generation

Google the hottest LinkedIn posts on "AI agents that browse the web", pull profiles of people who reacted/commented, and flag those matching "CTO building AI agents at a US startup".

Growth hacking
Atlassian Vercel Cursor Supabase Shopify Stripe

SaaS invoice collection

Collect every SaaS invoice each month, including the ones stuck behind a billing login, and email them to your accountant.

FinOps
Welcome to the Jungle Google Sheets

Candidate pipeline snapshot

List all current candidates for our dev position in our ATS, download their resumes, and generate a summary sheet for each one: target position, key experience, current status. Export to a gsheet.

HR

Start building not clicking

Get started for free