Real-time global intelligence dashboard. AI-powered news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational awareness interface
MCP server for Google Drive and Workspace
This server lets an MCP client talk to Google Drive, Docs, Sheets, Slides, and Calendar through one standardized interface. It supports search, file organization, document editing, sheet and slide creation, sharing, and calendar workflows while keeping Google credentials under your control.
Builders who want their agent to manage Google Drive and Workspace content through MCP.
You can ask your agent to find, edit, organize, and share Google files without leaving your client.
What it does
Drive and Workspace tools
Provides tools for file management, Shared Drives, permissions, revisions, Docs editing, Sheets formatting, Slides authoring, and Calendar events.
Multiple transport modes
Supports stdio for local clients, Streamable HTTP for hosted integrations, and team mode for shared deployments.
Identity control
Supports local OAuth, multiple Google accounts, per-tool account selection, service accounts, and externally managed OAuth tokens.
Optional resource interface
Exposes `gdrive:///` resources for direct reading and discovery alongside tools.
Self-hosted credentials
Keeps OAuth credentials and tokens in the environment you operate instead of a hosted service.
How to get it
- 1Download the JSON file and save it as
~/.config/google-drive-mcp/gcp-oauth.keys.json
- 2Run
npx -y @piotr-agier/google-drive-mcp auth
- 3Ask your client
Run authGetStatus and tell me which Google account is active.
- 4Then try a read-only Drive request
Search my Google Drive for files modified in the last seven days.
README
Google Drive MCP Server
Connect an MCP client to Google Drive, Docs, Sheets, Slides, and Calendar through one self-hosted server. Search and organize files, create and edit Workspace content, manage sharing, and automate multi-step workflows while keeping control of the Google identity and credentials used for every call.
Why this server
- Drive-first workflows: 116 tools cover file management, Shared Drives, permissions, revisions, rich Docs editing, Sheets formatting, Slides authoring, and Calendar events.
- Local or hosted: use stdio for a personal desktop client, Streamable HTTP for a hosted integration, or OAuth-protected team mode for a shared service.
- Identity control: local OAuth supports multiple Google accounts and per-tool account selection; service accounts and externally managed OAuth tokens are also supported.
- Agent-friendly access: tools expose targeted operations, while the optional
gdrive:///resource interface supports direct reading and discovery. - Open and self-hosted: credentials and tokens stay in the environment you operate.
This project remains focused on deep Drive and editor workflows rather than attempting to expose every Google Workspace API.
Client compatibility
Compatibility is determined by the transport and authentication flow a client supports.
| Client type | Transport | Recommended mode |
|---|---|---|
| Claude Desktop | stdio | Local OAuth |
| Other local MCP clients, including Gemini CLI | stdio | Local OAuth |
| claude.ai custom connectors | Streamable HTTP | Team mode |
| Other remote MCP clients | Streamable HTTP | Single identity behind access control, or team mode with OAuth 2.1 |
See Client configuration for configuration examples and transport requirements.
Quick start
1. Create Google OAuth credentials
In a Google Cloud project:
- Enable the Drive, Docs, Sheets, Slides, and Calendar APIs.
- Configure the OAuth consent screen and add your Google account as a test user when the app is in testing.
- Create an OAuth client with application type Desktop app.
- Download the JSON file and save it as:
~/.config/google-drive-mcp/gcp-oauth.keys.json
The setup guide has the complete Google Cloud walkthrough and alternative credential locations.
2. Authenticate
npx -y @piotr-agier/google-drive-mcp auth
Complete the Google consent flow in the browser. Tokens are stored by default at ~/.config/google-drive-mcp/tokens.json.
3. Add the server to your MCP client
For clients that use the common mcpServers configuration shape:
{
"mcpServers": {
"google-drive": {
"command": "npx",
"args": ["-y", "@piotr-agier/google-drive-mcp"]
}
}
}
Restart the client after saving its configuration. Claude Desktop paths and HTTP examples are documented in Client configuration.
4. Make a first tool call
Ask your client:
Run authGetStatus and tell me which Google account is active.
Then try a read-only Drive request:
Search my Google Drive for files modified in the last seven days.
If the identity is wrong or search returns no files, use the troubleshooting guide.
What you can do
Organize Drive
Find PDF files in /Reports, create an Archive folder there, and move files older
than one year into it.
Build a report
Create a Google Sheet for monthly results, summarize it in a Google Doc, and
create a short Google Slides presentation from the summary.
Edit a document
Find the "Project Plan" document, replace the old launch date, format the new
date in bold, and add a comment describing the change.
Prepare a meeting
Create a Calendar event with a Google Meet link and attach the project brief
from Drive.
Documentation
Every section that used to live in this README is now in one of these guides.
| Guide | Contents |
|---|---|
| Setup | Requirements, Google Cloud APIs, OAuth credentials, and installation |
| Client configuration | Supported transports and client configuration |
| Authentication | Local OAuth, multi-account, service accounts, external tokens, and scopes |
| Configuration | CLI flags, environment variables, defaults, and precedence |
| Deployment | Docker, Streamable HTTP, team mode, and reverse-proxy security |
| Tool reference | All Drive, Docs, Sheets, Slides, Calendar, and account tools |
| Troubleshooting | Authentication, API, identity, Docker, and rate-limit problems |
| Development | Repository structure, build commands, tests, and contributions |
Security
- Never commit OAuth credentials, service-account keys, access tokens, refresh tokens, or
tokens.json. - Use the narrowest OAuth scopes that support the tools you need.
- Keep the default HTTP bind address on
127.0.0.1unless the server is protected by TLS and access control. - Use team mode for shared deployments so every request is authenticated as its caller.
- Treat
team-store.jsonas a secret because it contains members' Google refresh tokens.
See Authentication and Deployment for the complete security and identity model.
Development and support
See the development guide to build and test the project.
- Report defects and request features in GitHub Issues.
- Review released changes in the changelog.
- Contributions are welcome through pull requests.
License
Acknowledgments
- Built on the Model Context Protocol.
- Uses the Google APIs Node.js Client.
Files in the repo
- .claude
- .github
- docs
- scripts
- src
- test
- .cursorignore
- .dockerignore
- .gitattributes
- .gitignore
- .npmignore
- CHANGELOG.md
- Dockerfile
- eslint.config.js
- gcp-oauth.keys.example.json
- LICENSE
- package-lock.json
- package.json
- README.md
- server.json
- tsconfig.json
- tsconfig.test.json
Discussion (0)
Ask about usage, or say what you built with itSign in to join the discussion.
No comments yet. Be the first to say what this is good for.
More connectors
High-performance code intelligence MCP server. Indexes codebases into a persistent knowledge graph — average repo in milliseconds. 158 languages, sub-ms queries, 99% fewer tokens. Single static binary, zero dependencies.

Universal provider proxy for OpenAI Codex & Claude Code — use any LLM (Claude, Gemini, Grok, DeepSeek, Ollama…) with Codex CLI, App, SDK, and Claude Code
Git-native persistent memory for AI coding agents. Implements Google OKF v0.2 with sub-300µs in-memory BM25 search, embedded MCP server, and progressive disclosure. Slashes token bloat by 80% with zero external databases or dependencies. Built in pure Go.
Local-first code intelligence graph for MCP and CLI. Builds a persistent map of your codebase so AI coding tools read only what matters, with benchmarked context reductions on reviews and large-repo workflows.
Stop your AI from making things up — it proposes, deterministic tools decide, every claim checked against ground truth with evidence. Grounded facts and context survive resets. Reverse engineering is the proving ground. MCP server + CLI.