PasteSheet icon PasteSheet logo mark — a spreadsheet grid with a curly brace on a green rounded square PasteSheet

Generate an MCP config for a Google Sheet

Every MCP client wants the same server described in a slightly different file, in a slightly different shape. Paste a Google Sheet URL and get the exact config block for each one, plus the path of the file it belongs in.

The sheet must be shared with "Anyone with the link". Nothing is stored.

Key facts

  • MCP is an open protocol built on JSON-RPC 2.0. A server exposes Resources, Prompts and Tools to any MCP client — Claude, Cursor, ChatGPT, VS Code and Windsurf all speak it. source
  • PasteSheet's MCP server exposes three tools — list_tabs, get_schema and query_rows — and is read-only by design: an agent can query your sheet but can never modify it.
  • It needs no Google Cloud project, no OAuth consent screen and no service-account JSON. You paste a share URL, and MCP is included on the Free plan.
  • The Google Sheets API allows 300 read requests per minute per project and 60 per minute per user. Past that it returns 429 RESOURCE_EXHAUSTED. source

What this generates

It checks that the sheet you pasted is readable, then writes out the MCP server config for Claude Desktop, Cursor, VS Code and Claude Code, and tells you which file each block goes in. The shapes genuinely differ — VS Code nests the server under a servers key, Cursor and Claude use mcpServers — and getting it wrong fails silently, with the client simply showing no tools.

Be clear about one thing: the config points at an endpoint URL, and that endpoint is created when you connect the sheet to PasteSheet. The generator fills in everything except the endpoint id, which is why the URL below carries a placeholder. Creating the endpoint is free and takes about a minute — no credit card, and MCP is on the Free plan.

What the AI can do once connected

The server exposes three read-only tools, and the agent decides when to call them:

  • list_tabs — discover the tabs in the sheet and which is the default.
  • get_schema — read each column's name, alias and inferred type before querying.
  • query_rows — read rows with exact-match filters, partial matches, full-text search, sorting and pagination.

It is read-only, and that is the point

The connection can query your sheet and can never modify it. There is no write tool to misuse and no scope that grants one, so handing the server to an agent is a bounded decision rather than a leap of faith.

It also needs no Google Cloud project, no OAuth consent screen and no service-account JSON. That is the entire setup burden of Google's own API, and it is the reason most people give up on connecting a spreadsheet to an agent. Here you paste a share URL.

Why an agent needs a cached endpoint

Agents are chatty. One exploring a question does not read a sheet once — it lists tabs, checks the schema, then fires several filtered queries in a row. Pointed straight at Google, that burns quota fast: the Sheets API allows 300 reads per minute per project and only 60 per minute per user before it starts returning 429.

A PasteSheet endpoint reads the sheet once, caches the rows, and serves every subsequent tool call from that cache — so the agent's tenth query costs Google nothing. See the MCP guide for the full walkthrough, or the Claude and Cursor guides for per-client steps.

What a live endpoint costs

Nothing, to start. The Free plan gives you 3 endpoints and 2,000 requests a month, with MCP included and no credit card. The tool above stays free and unlimited either way.

Free

For side projects and trying things out.

$0 /mo
Endpoints
3
Requests / mo
2,000
Row cap
500
  • MCP for AI agents

Frequently asked questions

Do I need a Google Cloud project or a service account?

No. PasteSheet reads sheets that are shared with anyone who has the link, so there is no Cloud project, no OAuth consent screen and no service-account JSON. That setup burden is the main reason people abandon connecting a spreadsheet to an AI agent.

Can the AI edit my spreadsheet?

No. The server exposes exactly three tools — list_tabs, get_schema and query_rows — and all three are read-only. There is no write tool to misuse and no scope that grants one, so the agent can analyse your data but never change it.

Which MCP clients does this work with?

Any MCP-compatible client. The generator writes the config for Claude Desktop, Claude Code, Cursor and VS Code, which cover most setups; ChatGPT connects through the account-wide workspace server using OAuth rather than a pasted config block.

Why does the URL contain a placeholder id?

Because the config points at an endpoint, and the endpoint is created when you connect the sheet. Everything else in the block is exactly what you need — only the id is filled in once the endpoint exists. Creating one is free, takes about a minute, and MCP is included on the Free plan.

Sources

Related tools

Turn your sheet into an API in minutes

Paste a Google Sheet URL and get a live REST API and MCP server — no backend, no code, free to start.