Reading, writing, restructuring and sharing are all here, which is unusual — most spreadsheet connections stop at reading. The counterweight is size: a tool list this long occupies context on every turn, and the filtering option exists precisely for that, letting you expose reading and writing alone when that is all the work needs.
A Google Sheets server covering the whole surface: finding spreadsheets in Drive, reading and writing cell ranges, adding rows, columns and sheets, copying and renaming, searching inside a spreadsheet, and sharing.
- Spreadsheets in a Drive folder listed, searched by name, and the folders themselves listed alongside
- Cell data read from one sheet, from several sheets at once, or several spreadsheets summarised together
- Formulas read back as formulas rather than as their computed results
- Cells updated, and batched updates applied across ranges in a single call
- Rows and columns added, and sheets created, renamed and copied
- A new spreadsheet created in a folder you name
- Text found inside a spreadsheet
- A spreadsheet shared with someone
- A way to enable only the tools you need, so the rest do not occupy the model's context
Google Cloud credentials with the Sheets and Drive APIs enabled. A service account is the recommended path — its key file and a Drive folder are supplied as environment settings — and OAuth is supported as well. Python, run with uv.
One command plus a key — uvx mcp-google-sheets@latest, then supply credentials