mcp-google-drive
Enables reading and searching Google Drive files, Google Docs, and Google Sheets via a CLI or MCP server, with support for section-based content extraction and Markdown import.
README
Google Drive CLI & MCP server
Read access to Google Drive, Docs, and Sheets through two surfaces that share
one implementation: a gdrive CLI (the primary surface, built for shell
pipelines and AI agents) and an MCP server (stdio). Both authenticate with
OAuth2 and call into a transport-agnostic core/ layer.
Requires Node.js >= 24.
This project includes code originally developed by Anthropic, PBC, licensed under the MIT License from this repo, and additional work from isaacphi/mcp-gdrive.
CLI
Build (npm run build), then invoke node dist/bin/gdrive.js <command>. For a
bare gdrive command, symlink dist/bin/gdrive.js into any directory on your
PATH. Every command accepts a Drive file ID or a Google URL.
gdrive search <query> [--type docs|sheets] [--page-size N] [--page-token TOKEN]
gdrive meta <fileId|url>
gdrive docs read <fileId|url> [--section "Heading"] [--json] [--no-images]
gdrive docs headings <fileId|url> [--min N] [--max N]
gdrive docs comments <fileId|url> [--include-resolved]
gdrive docs create --name "Title" [--from FILE] (or pipe markdown on stdin)
gdrive docs update <fileId|url> [--from FILE] (or pipe markdown on stdin)
gdrive sheets read <fileId|url> [--range A1:B10] [--csv|--tsv|--json]
gdrive files read <fileId|url>
docs create/update import Markdown via Drive's native converter. Updates
are a full-body replace: text and structure (headings, lists, tables,
links) convert, but it does not preserve embedded images or anchored
comments. Local image references () are stripped before import
(they make Drive's importer fail); public image URLs are kept and embedded.
Content is written to stdout so you can pipe or redirect it; structured
output is available via --json, and auxiliary notes (e.g. the embedded-image
manifest) go to stderr. Binary content is refused on an interactive terminal —
redirect it (gdrive files read <id> > out.pdf). Run any command with
--help for its flags.
Examples:
gdrive docs read <id> > prd.md # whole doc as Markdown
gdrive docs read <id> --section "2. Context" # just one section
gdrive sheets read <id> --range "'Sheet2'!A1:C9" --tsv
gdrive search "quarterly" --type docs
MCP Components
Tools
-
gdrive_search
- Description: Search for files in Google Drive.
- Input:
query(string): Search query.pageToken(string, optional): Token for the next page of results.pageSize(number, optional): Number of results per page (max 100).
- Output: Returns structured JSON with file metadata by default. Set
MCP_GDRIVE_OUTPUT_FORMAT=textto return a plain text list.
-
gdrive_read_file
- Description: Convenience tool to read contents of a file from Google Drive (legacy behavior preserved).
- Input:
fileId(string): ID of the file to read.url(string, optional): Google Docs/Drive URL with heading anchor (for section extraction).sectionHeading(string, optional): Heading text to extract a section.
- Output: Returns structured JSON with file metadata and content by default, or plain text with
MCP_GDRIVE_OUTPUT_FORMAT=text.
-
gdrive_parse_link
- Description: Parse a Google Docs URL and extract identifiers.
- Input:
url(string): Google Docs URL to parse.
- Output: Returns
fileId, optionalheadingId, anddocType.
-
gdrive_get_metadata
- Description: Fetch file metadata and optional headings for Google Docs.
- Input:
fileId(string): ID of the file to inspect.includeHeadings(boolean, optional): Include headings when file is a Google Doc.
- Output: Returns file metadata, optional
docType, and optional headings list.
-
gdrive_list_headings
- Description: List headings for a Google Doc.
- Input:
fileId(string): ID of the Google Doc to inspect.minLevel(number, optional): Minimum heading level to include (e.g.,2for H2+).maxLevel(number, optional): Maximum heading level to include (e.g.,3for up to H3).
- Output: Returns a headings list (with levels) and file metadata.
-
gdrive_read_content
- Description: Read content explicitly by mode (
fullorsection). - Input:
fileId(string): ID of the file to read.mode(string, optional):full(default) orsection.sectionHeading(string, optional): Required whenmode="section".
- Output: Returns file metadata and content, or a specific section when requested.
- Description: Read content explicitly by mode (
-
gdrive_download
- Description: Download content to a local file, returning byte offsets for local paging.
- Input:
fileId(string): ID of the file to read.mode(string, optional):full(default) orsection.sectionHeading(string, optional): Required whenmode="section".destinationPath(string, optional): Output path or directory for the download.chunkSizeBytes(number, optional): Chunk size for offsets in bytes.
- Output: Returns file metadata, download path, and byte offsets for local paging.
Download directory
The default download directory can be configured via GDRIVE_DOWNLOAD_DIR. If not set, downloads go to:
~/.mcp-gdrive/downloads
Output format
Tool output format is configured globally via environment variable:
MCP_GDRIVE_OUTPUT_FORMAT=json # default
MCP_GDRIVE_OUTPUT_FORMAT=text
Agent Workflow (Docs Link)
- Parse the link with
gdrive_parse_linkto getfileIdandheadingId. - List headings with
gdrive_list_headings(usegdrive_get_metadatafor file metadata). - Read with
gdrive_read_contentusingmode="section"per heading (avoid full reads for large docs).
Prompts
The server exposes prompt templates to guide agents through common workflows:
outline_doc(args:url, optionalminLevel,maxLevel)read_section_by_heading(args:url,sectionHeading)
Sheets
The MCP server reads spreadsheets as CSV via the generic read tools. For
range- and tab-aware reads (the Sheets values API), use the CLI's
gdrive sheets read. Write operations (creating/updating Docs or Sheets) are
not implemented on either surface.
Resources
The server does not currently expose any gdrive:/// resources. Access Drive
content via tools.
Setup
Three steps, ~3 minutes. You never edit a .env — credentials are read from
the file you download from Google.
1. Install
git clone https://github.com/abanka27/mcp-google-drive.git
cd mcp-google-drive
npm install # builds automatically (requires Node >= 24)
2. Get Google credentials (one-time, ~2 min)
Each person needs their own OAuth client — credentials can't be shared. Click through these exact pages:
- Create or pick a project: https://console.cloud.google.com/projectcreate
- Enable three APIs — click Enable on each:
- Drive — https://console.cloud.google.com/apis/library/drive.googleapis.com
- Docs — https://console.cloud.google.com/apis/library/docs.googleapis.com
- Sheets — https://console.cloud.google.com/apis/library/sheets.googleapis.com
- OAuth consent screen: https://console.cloud.google.com/apis/credentials/consent Pick Internal if your account offers it. (An External app in "Testing" expires logins every 7 days — add yourself as a test user, or publish it.)
- Create credentials → OAuth client ID, application type Desktop app: https://console.cloud.google.com/apis/credentials/oauthclient
- Download the JSON. That single file is all you need.
3. Run setup
npm run setup
This finds the file you just downloaded (in the current directory or
~/Downloads), stores it, opens your browser to sign in once, and links
gdrive onto your PATH. If it can't find the file, pass the path explicitly:
node dist/bin/gdrive.js setup ~/Downloads/client_secret_XXXX.json
Done — try it:
gdrive search "" --page-size 3
Credentials and the saved token live in ~/.config/mcp-gdrive
($XDG_CONFIG_HOME/mcp-gdrive), or set GDRIVE_CREDS_DIR to override. The
browser prompt only appears on first run; after that the token refreshes
silently.
Use with an AI agent (Claude Code skill)
This repo ships a skill at .claude/skills/gdrive/ that teaches an agent the
commands and when to reach for them. It loads automatically when you run Claude
Code inside this repo. To use it from any directory, copy it into your user
skills:
cp -r .claude/skills/gdrive ~/.claude/skills/gdrive
Then ask your agent to read, search, or update Google Docs/Sheets and it will
drive gdrive for you.
License
This MCP server is licensed under the MIT License. This means you are free to use, modify, and distribute the software, subject to the terms and conditions of the MIT License. For more details, please see the LICENSE file in the project repository.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.