Skip to content

Decision: Fork and Continue as Independent Project

Date: 2026-05-09 Snapshot commit: 9d4df64 — codebase as it existed when this decision was made

This is a point-in-time record. It captures context, alternatives, and reasoning as they were understood on the date above — not the current state of the project.

Background

This project began as a fork of xing5/mcp-google-sheets, a community MCP server for interacting with Google Sheets. The original repo was chosen specifically because it included get_sheet_formulas — access to raw cell formulas via the Sheets API — which most alternatives lacked. As of May 2026, Google has not released an official MCP server for Sheets or Drive (official servers exist for Maps, BigQuery, GCE, and GKE), so community solutions are the only option.

Over time the codebase diverged substantially. It now includes a caching architecture, Google Drive file listing, Google Docs support, tool filtering, and observability improvements that put it well outside the original scope. This document records the research done to decide whether to continue building on this base or switch to a different foundation.

Alternatives Evaluated (May 2026)

xing5/mcp-google-sheets — current upstream

  • Language: Python
  • Tools: ~19 tools covering core Sheets read/write, formulas, batch operations, Drive spreadsheet management
  • Strengths: Deep Sheets API coverage, get_sheet_formulas, raw batch_update passthrough, active maintenance (last commit March 2026)
  • Weaknesses: No cell formatting, no Drive file management beyond spreadsheets, no Docs support

freema/mcp-gsheets

  • Language: TypeScript
  • Tools: 40+ tools — the most comprehensive Sheets-specific server found
  • Strengths: Cell formatting (colors, fonts, borders, number formats), conditional formatting, data validation, sheet dimension metadata, freeze rows/columns, compact formatting snapshots
  • Weaknesses: No get_sheet_formulas, no Docs API support, TypeScript (harder to extend for this use case), no caching layer
  • Notable: Primary inspiration for the feature roadmap

piotr-agier/google-drive-mcp

  • Language: TypeScript
  • Tools: Cross-service Workspace suite (Sheets, Docs, Slides, Calendar)
  • Strengths: Named ranges, protected ranges, Drive file operations (move, copy, delete, rename), permission management, revision history
  • Weaknesses: Shallower Sheets depth than xing5 or freema, no formula access, TypeScript
  • Notable: Primary inspiration for Drive file operations and permissions roadmap items

isaacphi/mcp-gdrive

  • Language: TypeScript
  • Tools: 4 tools (search, read_file, gsheets_read, gsheets_update_cell)
  • Verdict: Too minimal to consider as a base

Decision: Continue and Fork

Continue building on the xing5 base. No alternative offers a clearly superior starting point:

  • freema/mcp-gsheets has broader formatting coverage but lacks formula access (the original selection criterion) and is TypeScript
  • piotr-agier/google-drive-mcp has better Drive operations but shallower Sheets depth
  • Neither has a caching layer
  • No official Google-provided alternative exists

The gap analysis (see roadmap.md) shows that most missing features can be added incrementally using the existing architecture. The batch_update passthrough already provides an escape hatch for any Sheets API operation not yet wrapped as a named tool.

Fork from upstream. The codebase has grown well past the original scope. Continuing as a silent fork accumulates invisible divergence and makes it harder to credit the work appropriately. The plan:

  1. Open a PR upstream from the upstream-observability branch — structured logging, per-tool timing, and cache_discovery=False fix — changes that fit cleanly within xing5's scope before cutting loose
  2. Fork to a new repo under an appropriate name (e.g., mcp-google-workspace)
  3. Update README to credit xing5 as the original base and freema/piotr-agier for roadmap inspiration