Open source Google Sheets MCP servers, compared

What people mean by an open source Sheets MCP server

MCP is an open protocol — the specification itself is public, and anyone can implement a server against it. So in one sense every MCP server speaks an open standard. But when people search for an open source Google Sheets MCP server, they usually mean something narrower: source code they can read, fork, audit, and run on their own machine, with no vendor in the request path.

That is a legitimate requirement, and it is worth being clear about which servers actually meet it — and what the day-two cost of running one looks like once the demo works.

The options that are genuinely open source

Two categories of code are worth knowing about:

• Community servers such as mcp-google-sheets — permissively licensed, installable via uvx or Docker, and typically read-write. You supply the Google credentials and the process.
• Google's own Workspace MCP server — Google publishes an open-source Workspace MCP server as a Gemini CLI extension, and that one does include Sheets. But it is the self-hosted flavour: Google's hosted MCP servers skip Sheets entirely.
• Roll your own — the protocol is JSON-RPC 2.0 and the Sheets API is documented, so a minimal server is a weekend. Keeping it correct, cached, and patched is the part that is not a weekend.

What running one actually costs

None of these servers can read a sheet until Google trusts them. That means a Google Cloud project, the Sheets API enabled, and either OAuth credentials or a service-account JSON key — the last of which is a long-lived secret you now have to store, rotate, and keep out of your shell history.

Then there is the process itself. An MCP server is a running service: something has to start it, restart it when it dies, patch it when a dependency ships a CVE, and keep it reachable from whichever client you are using. And because most community servers default to read-write, a confused agent can modify the spreadsheet — which is exactly the failure mode you do not want to discover in production.

The licence is free. The Cloud project, the secret rotation, the uptime, and the blast radius are the actual price.

Self-hosted open source vs a hosted endpoint

Neither column is strictly better — they optimise for different things:

Is PasteSheet open source?

No — and it would be dishonest to imply otherwise on a page about open source. PasteSheet is a hosted service. If your requirement is auditable source code or data residency on your own hardware, a community server or Google's self-hosted Workspace extension is the right answer, and this page is not trying to talk you out of it.

Where PasteSheet wins is the other requirement: you want an agent reading a spreadsheet today, you do not want a Cloud project, and you want the connection to be structurally incapable of writing. That is a different trade, not a better one. If self-hosting is the sticking point rather than the source code, the hosted alternative guide covers the same ground from that angle.

What the hosted route 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.

Frequently asked questions

Sources

• Model Context Protocol specification — Model Context Protocol
• google/mcp — Google's open-source MCP servers — Google
• Authorize requests — Google Sheets API — Google
• Service accounts overview — Google Cloud

Related guides