google-sheets-mcp-and-skill

# google-sheets-mcp-and-skill [](https://github.com/colindmurray/google-sheets-mcp-and-skill/actions/workflows/ci.yml) [](https://pypi.org/project/google-sheets-mcp-and-skill/) [](https://pypi.org/project/google-sheets-mcp-and-skill/) Read a Google Sheet the way it actually works — formulas, real colors, conditional-format rules, native tables, filter state, in-cell rich text, and the comments humans left on it — then write it back safely, or export the whole thing to PDF, Excel, or CSV. Render any read as text, JSON, JSONL, CSV, TSV, or Markdown. One core library, exposed as both an MCP server and a CLI (with a bundled skill). ## The problem A serious spreadsheet is rarely just a grid of values. It's a small application: derived columns built from formulas, cells that change color from conditional-format rules, merged headers, dropdowns, native tables with typed columns, filter views that hide half the rows, frozen panes. The values you can see are the *output*; the logic that produces them lives in the formulas, the formatting, the conditional-format rules, and the structure around them. To understand such a sheet before changing it, whether you're an AI agent or a person, you have to read that logic losslessly. A screenshot tells you a cell is red, but not *why* it's red, what formula feeds it, which rule would turn it green, or that the table is filtered so the row you're about to edit isn't the row you're looking at. "What's the value in C7?" is the easy question. "What formula computes C7, which conditional-format rule sets its background and at what threshold, and is this column part of a named table?" is the one that matters, and the one most tooling can't answer. That gap is the whole point of this project. The…