Docs: System guides

Read Google Sheets into Maplith

Google Sheets holds the lists a company keeps outside its systems (price lists, order trackers, stock counts, budgets), which often fill the gaps between the real systems.

Spreadsheet · Also known as Google Drive spreadsheets, Google Workspace Sheets · Checked 2026-10-08

On this page

Best path

Recommended: your agent fetches it. Most business sheets are a few hundred to a few thousand rows, and your agent can usually already read them through its Google Drive connection. Google's own Sheets MCP server is in Developer Preview only, so it is not yet a path most customers have.

Try the paths in this order:

Your agent fetches it, with the access you already gave it. Connectors a vendor or client page confirms: Google Drive connector in Claude; Google Drive app in ChatGPT. Your agent uses its Google Drive connection to find the sheets that matter, reads the header row and a sample of rows from each tab, and never edits them.
Maplith reads an MCP server you already run. Google Sheets runs Google Sheets MCP server. Its address comes from Google Sheets’s own setup instructions. Documentation: developers.google.com. The server signs in with Google OAuth. Set up the OAuth client with only the read-only scopes (spreadsheets.readonly and drive.readonly) so the token can read but not change sheets. Available only through the Google Workspace Developer Preview Program and needs a Google Cloud project; it also has tools that write cells, formulas and rows, so use read-only scopes. Maplith cannot read this server itself, because it offers tools or a personal sign-in to an agent. Connect it to your agent instead, as the first path.
Export files, the fallback that always works. See the export steps below.

Datasets that make a good map

Dataset	One row is	Key columns	Joins to
A sheet tab used as a table	One row per record (an order, a product, a customer, a booking) with one header row on top.	a stable id column (order number, SKU, customer number), date, status, amount	The system the list shadows, by the same id: shop orders by order number, products by SKU, accounting customers by customer number.
Price or product list	One row per product or service with its price and attributes.	SKU or product number, product name, unit price, currency	Shop products and order lines by SKU; accounting products by product number.
Manual tracker (orders, tasks, cases)	One row per item someone follows up by hand.	reference id from another system, owner, status, date updated, customer email (Maplith tries to mask it)	CRM deals or helpdesk tickets by their id or by customer email.
Budget or KPI sheet	One row per period and line, often typed in by hand.	period, account or metric name, value	Accounting accounts by account number; ad spend and revenue by month.

Columns are named, never their values. Maplith tries to mask personal data such as email addresses, but that is best effort: you decide what goes in.

The access to give

Who usually has the rights: The person who owns the sheet; in a company Google Workspace, an admin can see who owns which files.

Choose the narrowest access that covers the data:

Viewer
spreadsheets.readonly
drive.readonly

Where to click:

Open Google Drive and select the sheet
Click Share
Add the person or account that needs access
Choose the Viewer role
Click Send or Share

A key or token made here stays in the tool that reads the system, never in a chat. Maplith asks for one only under Data sources › Systems in the app.

Export steps

Format: CSV Date range: The whole sheet (sheets rarely have a date filter)

Open the sheet and click the tab you want
Click File, then Download
Choose Comma-separated values (.csv)
Repeat for each tab that holds a table
A sheet is easiest to map when each tab has one header row, one table and a stable id column, with no merged cells or totals rows inside the table

How much data

About 50–2,000 new rows across the sheets that matter a month (an estimate). Estimate for a small company, not a measured figure: sheets are side lists and trackers, not the main system of record.

An agent that reads rows through a chat reads every one of them, about one token per four characters. Above a few thousand rows a dataset, or for a daily refresh, an MCP server or an export is the better path.

Always

Read-only, and never a credential sent to Maplith through a chat.
The same recent period, a few months, from every system.
You decide what goes in.

Sources

Written from public documentation, last checked 2026-10-08:

← All system guides · Markdown version