Skip to content

mcp-gee-sweet

An MCP server that gives AI clients reliable, direct access to Google Workspace — Sheets, Drive, Docs, and Calendar.

What it does

  • Read and write Google Sheets — cell values, formulas, sheet structure, batch updates
  • Manage Google Drive — list, search, upload, download, move, copy, and sync files and folders
  • Work with Google Docs — create documents, read content, write formatted content from HTML
  • Access Google Calendar — list calendars and events, create and update events, find free slots

Why it exists

This project started as a fork of xing5/mcp-google-sheets and has since diverged substantially — full Sheets read/write/formulas/structure/formatting/charts plus a raw batch_update passthrough for anything a named tool doesn't cover, an AST-based Docs pipeline (HTML and Markdown input, theming, structural table editing), Drive file/sharing/sync operations, and Calendar. See Project Fork for the original fork history.

The MCP landscape for Google Workspace has grown since then — Google now runs an official Developer Preview program, and several community projects cover similar ground. Rather than claim to be the only option, here's how the coverage actually breaks down as of mid-2026 (full sourcing in Repositioning):

mcp-gee-sweet Official Google Workspace MCP (Developer Preview) Notable community alternatives
Sheets 35 tools — read/write, formulas, structure, formatting, charts, raw batch_update passthrough 6 tools (Developer Preview) taylorwilsdon/google_workspace_mcp (12), a-bonus/google-docs-mcp (~44, incl. comments/conditional formatting/protected ranges we don't have yet)
Docs 34 tools — HTML + Markdown input, Markdown export, AST-based theming (apply_theme, named styles), structural table editing 2 tools (Developer Preview) taylorwilsdon (18), piotr-agier/google-drive-mcp (~17, incl. footnotes/multi-tab we don't have)
Drive 38 tools 8 tools Several — piotr-agier and others cover a similar range
Calendar 17 tools 9 tools Several
Release channel Stable PyPI releases behind a QA regression gate Developer Preview Varies by project

This project deliberately keeps its named-tool surface smaller than some alternatives — a-bonus/google-docs-mcp, for example, registers more raw tools by giving nearly every Sheets/Docs operation its own dedicated tool. Here, related operations consolidate into fewer, more parameterized tools, with batch_update as the escape hatch for whatever doesn't fit a named tool — see Design Principles for the reasoning ("tool count is a cost"). Neither approach is strictly better; it's a different tradeoff, and it's why raw tool counts alone don't tell the whole story in the table above.

Known gaps found while researching this comparison (comments, conditional formatting, protected ranges, multi-tab Docs, and more) are tracked as candidates in the roadmap, not silently ignored.

Key design decisions

Tool count is a cost. Every registered tool is a name the AI must reason about on every call. This server applies a strict inclusion test — see Design Principles — and defers borderline cases to the batch_update passthrough rather than adding named tools speculatively.

Composite tools only when it matters. Multi-step workflows are left to the AI client unless they involve binary data, pagination loops, or encoding decisions that Claude handles inconsistently in practice. See Design Principles for the policy.

Caching by default. Sheet structure, sheet data, Drive folder listings, doc content, and calendar metadata are all cached in a local SQLite database with TTL + dirty invalidation. See Configuration for details.

Quickstart

# Run locally
uv run mcp-gee-sweet

# Docker
docker run --rm -p 8000:8000 \
  -e CREDENTIALS_CONFIG=<base64-service-account> \
  -e DRIVE_FOLDER_ID=<folder-id> \
  mcp-gee-sweet

Reference