# Connect Google Sheets to an n8n AI agent n8n already has a Google Sheets node. So why point an n8n AI Agent at an MCP server instead? Because a node fetches what you told it to — an agent decides what it needs. *Last updated: 2026-07-13 · Source: * ## Key facts - n8n ships an **MCP Client Tool** node you attach to an AI Agent node. Point it at a server URL and the agent can call that server's tools while the workflow runs. ([source](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.toolmcp/)) - 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](https://developers.google.com/workspace/sheets/api/limits)) - A cached endpoint reads the sheet **once per TTL**, not once per request — so 10,000 visitors become one upstream read, and Google's quota stops being your problem. - 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. ## Why use MCP when n8n has a Sheets node? The built-in Google Sheets node is the right tool when *you* know exactly which rows you want — it is deterministic, and it can write. Use it for that. An MCP server is for when the **agent** decides. Give an n8n AI Agent node a [Google Sheets MCP server](https://pastesheet.com/guides/google-sheets-mcp) and it can read the schema, choose a filter, and fetch only the rows a user's question actually needs — without you hard-wiring the query in advance. ## Wire it up 1. Add an **AI Agent** node to your workflow. 2. Add an **MCP Client** tool to that agent. 3. Set the server URL to your endpoint's MCP URL from the **Connect via MCP** panel. 4. For a private endpoint, send `Authorization: Bearer YOUR_KEY` as a header. 5. Prompt the agent normally — it will call `get_schema` and `query_rows` on its own. ## Why not point the agent at Google directly? Because agents retry, explore, and ask follow-up questions. Google allows **300 reads per minute per project and 60 per minute per user** before returning a `429` ([published limits](https://developers.google.com/workspace/sheets/api/limits)). A cached endpoint absorbs that: the sheet is read once, and every subsequent tool call is served from cache. ## What it costs **MCP is included on the Free plan.** Connect any **public** endpoint over MCP with no Google Cloud project and no credit card. Private endpoints and the account-wide workspace server need a paid plan (from **Starter ($9/mo)**) for the keys and OAuth they authenticate with. Pro adds full-text search and aggregation (`count`, `sum`, `avg`, `group_by`) that your AI agent can call through `query_rows`. **Free — $0/month.** For side projects and trying things out. - Endpoints: 3 - Requests: 2,000 / month - Rows per endpoint: 500 - Tabs per endpoint: 1 - Rate limit: 60 / minute ## Frequently asked questions ### Can n8n connect to an MCP server? Yes. n8n provides an MCP Client tool you attach to an AI Agent node. Point it at an MCP server URL and the agent can call that server's tools during a workflow run. ### Should I use the Google Sheets node or MCP? Use the node when you know the exact rows you want, or when you need to write. Use MCP when an AI agent should decide what to fetch based on a question it was asked. ### Can the n8n agent write back to my sheet? Not through PasteSheet — it is read-only. If your workflow needs to write, use n8n's native Google Sheets node for that step and MCP for the reading and reasoning. ### Does this work in self-hosted n8n? Yes. The MCP server is a hosted HTTPS URL, so any n8n instance that can reach the internet can use it — cloud or self-hosted. ## Sources - [MCP Client Tool node](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.toolmcp/) — n8n - [Usage limits — Google Sheets API](https://developers.google.com/workspace/sheets/api/limits) — Google ## Related guides - [Google Sheets MCP Server for Claude & ChatGPT](https://pastesheet.com/guides/google-sheets-mcp) — Turn any public or private Google Sheet into an MCP server so Claude, Cursor, and ChatGPT can read and query it in plain English. No code, no backend. - [Connect Google Sheets to Zapier & Make](https://pastesheet.com/guides/google-sheets-api-zapier-make) — Feed Google Sheets data into Zapier or Make as clean JSON. Expose your sheet as a REST API, then read it from any no-code automation — no add-on. - [Google Sheets MCP vs Zapier & Composio](https://pastesheet.com/guides/google-sheets-mcp-vs-zapier) — Zapier and Composio give Google Sheets MCP for read-write automation. PasteSheet is a hosted, read-only endpoint from a pasted URL — which to pick, and when. - [Read-Only Google Sheets MCP for Agents](https://pastesheet.com/guides/read-only-google-sheets-mcp) — A read-only Google Sheets MCP server is safe to hand an AI agent: it can query, filter, and analyze your data but can never modify the spreadsheet. --- [PasteSheet](https://pastesheet.com) turns any Google Sheet into a live REST API and MCP server for AI agents — no backend, no code. Canonical HTML version of this page: