Listed from the MCP registry; nobody on Odel publishes it.
MCP server for the Google Docs API: read and edit documents, styles, tables, images, comments.
Runs on your machine from its package; the steps are below.

At most: creates or changes data: records, files, settings.
Odel charges nothing per call. The publisher may charge for its own service.
Runs locally over stdio
This server isn't hosted — your MCP client launches it from a package registry. Use one of the commands below, or drop the config into your client (e.g. Claude Desktop).
Run
npx -y @a1-x-tech/mcp-google-docs
MCP client config
{
"mcpServers": {
"google-docs-mcp": {
"command": "npx",
"args": [
"-y",
"@a1-x-tech/mcp-google-docs"
]
}
}
}English | Русский
A1 Google Docs MCP lets an AI app read and edit Google Docs in plain language. Read a document as text or Markdown, change an exact passage, style headings, lists and tables, work through comment threads and export the result to PDF or DOCX.
It uses the Google Docs API with your Google account. It edits by exact index ranges rather than by guesswork, and makes the limits of the Docs API explicit instead of implying that every document task is possible.
127.0.0.1 with PKCE and keeps the tokens itself — no config files, no restart.Start with a read-only question:
Read the launch plan document and summarize its unresolved comment threads.
Connect the server · Explore use cases · Open technical documentation
You: Show me the text and comments of the launch plan document.
Assistant: Reads the document as compact text blocks and lists its comment threads. Nothing changes.
You: Rewrite the “Timeline” paragraph to say the beta starts on March 3.
Assistant: Shows the exact range it will replace and the proposed text, then asks for confirmation before editing.
You: Confirm.
Assistant: Replaces that one range. The rest of the document, its formatting and its comments stay as they were.
You need Node.js 20+ and a Google account. Credentials are not required at install time — the server connects from the conversation.
In the app: open Settings → MCP servers, select Add server, choose STDIO, enter the command npx -y @a1-x-tech/mcp-google-docs@latest and environment variables GOOGLE_DOCS_CLIENT_ID, GOOGLE_DOCS_CLIENT_SECRET, GOOGLE_DOCS_REFRESH_TOKEN, then select Save and Restart.
From the command line:
codex mcp add google-docs \
-- npx -y @a1-x-tech/mcp-google-docs@latest
codex mcp list
claude mcp add \
--transport stdio --scope user google-docs \
-- npx -y @a1-x-tech/mcp-google-docs@latest
claude mcp list
The current official path is Settings → Extensions. For a custom desktop extension, open Advanced settings → Extension Developer → Install Extension…, select a .mcpb file and follow the prompts.
This repository currently publishes an npm stdio package and does not contain a .mcpb bundle. For Claude Desktop builds that still support local configuration, use the following JSON stdio configuration as a fallback:
{
"mcpServers": {
"google-docs": {
"command": "npx",
"args": ["-y", "@a1-x-tech/mcp-google-docs@latest"]
}
}
}
In those builds, save it to ~/Library/Application Support/Claude/claude_desktop_config.json on macOS or %APPDATA%\Claude\claude_desktop_config.json on Windows.
Add this to ~/.cursor/mcp.json on macOS/Linux or %USERPROFILE%\.cursor\mcp.json on Windows:
{
"mcpServers": {
"google-docs": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@a1-x-tech/mcp-google-docs@latest"]
}
}
}
Run MCP: Open User Configuration and add:
{
"servers": {
"google-docs": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@a1-x-tech/mcp-google-docs@latest"]
}
}
}
Check it with MCP: List Servers.
create_document creates a document — empty, or converted from Markdown in one call.read_document_text before each edit and to edit from the end of the document backwards.import_markdown replaces the entire body: comment anchors, positioned objects, headers/footers and extra tabs do not survive the conversion.Exports are capped at 10 MB and do not include comments or suggestions. Inline images are fetched by Google from a public URL (PNG/JPEG/GIF, up to 50 MB and 25 megapixels); there is no channel for uploading image files.
| Operation | What happens | Confirmation boundary |
|---|---|---|
| Read a document, its tabs or comments | Reads content and structure | No change |
| Export a document | Writes a local file when output_path is set; the document itself is untouched | Changes local files only |
| Create a document | Adds a new document | Changes Google Docs |
| Insert text, a table, a break or an image | Adds content | Changes a document |
| Style text or paragraphs, manage bullets | Overwrites formatting on a range | Changes a document |
| Replace or delete a range, find and replace | Removes existing content | Destructive |
| Replace the whole document with Markdown | Replaces the entire body | Destructive |
| Manage comments | Creates, replies, resolves or permanently deletes | Potentially destructive |
| Raw API request | Can call API methods without a dedicated tool | Potentially destructive |
The AI client controls confirmation prompts. The server marks reads, writes and destructive tools so the client can distinguish an inspection from a live change.
Google Docs requires OAuth 2.0; an API key is not enough. There are two ways in, and the first one needs no configuration files.
Say "connect Google Docs" and the assistant runs the flow with you:
setup_instructions prints the checklist: create or select a Google Cloud project, enable Google Docs API, configure the consent screen and create a Desktop app OAuth client.set_client stores it owner-only. The secret never goes through the conversation.start_login returns a Google consent link. Open it on this machine and approve; the code comes back to a one-shot listener on 127.0.0.1 (PKCE), never through the chat.finish_login exchanges the code and saves the tokens to ~/.config/mcp-google-docs/credentials.json (mode 0600).The tokens are re-read on every call, so the connection works immediately — no restart of the AI app. auth_status shows what is connected, logout revokes and deletes it.
Create or select a Google Cloud project and enable both the Google Docs API and the Google Drive API (export, Markdown conversion and comments go through Drive endpoints).
Configure the OAuth consent screen and create a Desktop app OAuth client.
Authorize the Google account that owns or can edit the documents. The OAuth 2.0 Playground can obtain the refresh token when Use your own OAuth credentials is enabled.
Request both scopes:
https://www.googleapis.com/auth/documents
https://www.googleapis.com/auth/drive
For a narrower setup, drive.file is enough when export, Markdown and comments only touch documents created through this OAuth client, and the read-only pair documents.readonly + drive.readonly is enough for the read-only tools.
Testing-mode OAuth refresh tokens can expire after seven days. Publish the OAuth app, or use an Internal app in a Workspace domain, when you need long-lived access. Treat the client secret and refresh token as passwords.
Every variable is optional — with none of them the server connects from the chat.
| Variable | Required | Description |
|---|---|---|
GOOGLE_DOCS_CLIENT_ID | No* | OAuth client ID. |
GOOGLE_DOCS_CLIENT_SECRET | No* | OAuth client secret. |
GOOGLE_DOCS_REFRESH_TOKEN | No* | OAuth refresh token. |
GOOGLE_DOCS_ACCESS_TOKEN | No* | Short-lived alternative to the OAuth trio (~1 hour). |
GOOGLE_DOCS_OAUTH_PORT | No | Fixed loopback port for the in-chat login; useful over SSH port forwarding. |
GOOGLE_DOCS_API_BASE | No | Google Docs API base URL override. |
GOOGLE_DOCS_DRIVE_API_BASE | No | Drive API base URL override (export, Markdown, comments). |
GOOGLE_DOCS_TIMEOUT_MS | No | Per-request timeout; default 60000 ms. |
GOOGLE_DOCS_MAX_RETRIES | No | Temporary-error retries; default 3. |
* Provide either the OAuth trio or an access token.
#; a value from your request can appear in it if the error text contains it. Set ASKADS_TELEMETRY=0 to opt out.429, the server backs off and retries; reads also retry after network and 5xx errors, while writes are never replayed after an uncertain failure — a replayed write could duplicate the edit.Found a bug or need a scenario? Create an issue or write in Telegram.
You made it to the end!