Search and read your Drive — Docs, Sheets and files become context your assistant can actually use.
Most organisational knowledge lives in somebody's Drive, in a folder structure nobody fully remembers. The Google Drive MCP server makes that searchable and readable from your assistant: query across your files, open the hits, and work with their contents — with Google's native formats converted automatically (Docs to markdown, Sheets to CSV) so the model gets clean text rather than proprietary blobs.
The pattern it unlocks is "answer from our documents." Point a question at your Drive — what did we decide about pricing in last quarter's planning doc? what's in the onboarding checklist? — and the assistant searches, reads the relevant files and answers with the actual content in context. For meeting-notes archaeology and policy lookups, that beats fifteen minutes of manual Drive searching every time.
Setup is the honest cost of entry. This is a Google Workspace integration, so you'll create a Google Cloud project, enable the Drive API, configure an OAuth consent screen and complete a browser auth flow the first time. None of it is hard, but it's fifteen minutes of console-clicking that the one-line installs in this directory don't need. The token refreshes itself afterwards.
Access is read-only in the reference implementation — search and read, no writing or sharing changes — which is the right default for pointing an AI at a corpus of company documents. Like its reference siblings it's now archived (community forks add Sheets editing and broader Workspace coverage if you need writes), but for the core "let my assistant read our Drive" job, it remains the straightforward answer.
auth command on its own once. It opens a browser and writes the credentials JSON that GDRIVE_CREDENTIALS_PATH points at; skip it and you'll be debugging a startup crash instead.Questions answered with the actual contents of planning docs, specs and notes.
"What did we decide about X?" resolved from months of scattered documents.
Sheets arrive as CSV, ready for summarising or analysis.
| Tool | What it does |
|---|---|
| search | Full-text search across your Drive files |
| gdrive_read_file | Read a file's contents, auto-converting Google formats (Docs→markdown, Sheets→CSV) |
{
"mcpServers": {
"gdrive": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-gdrive"],
"env": {
"GDRIVE_CREDENTIALS_PATH": "/path/to/.gdrive-server-credentials.json"
}
}
}
}Requires a one-time OAuth setup: create a Google Cloud project, enable the Drive API, download OAuth keys, then run the server's auth command.
A Google account, a Google Cloud project with the Drive API enabled, and a one-time browser OAuth flow.
| Variable | Description | Required |
|---|---|---|
| GDRIVE_CREDENTIALS_PATH | Path to the saved OAuth credentials produced by the auth flow | Yes |
| GDRIVE_OAUTH_PATH | Path to the OAuth client keys JSON downloaded from Google Cloud | Optional |
Scoped local file access — read, write, search and reorganise files in directories you explicitly allow.
Query, modify and analyse local SQLite databases in conversation — the fastest way to chat with a data file.
Chat with your second brain — search, read and write vault notes through the Local REST API.
Model Context Protocol (MCP) tool to interact with Claude Desktop on macOS
Upload, search, transform and organise your Cloudinary media library from a chat window
Run an M&A data room — projects, folders, permissions and diligence Q&A — from an AI conversation.