Google Workspace MCP Server
@taylorwilsdon
基本情報
Google Workspace MCP Server について
Control Gmail, Google Calendar, Docs, Sheets, Slides, Chat, Forms, Tasks, Search & Drive with AI - Comprehensive Google Workspace / G Suite MCP Server & CLI Tool
Google Workspace MCP Server の設定
以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。
{
"mcpServers": {
"google_workspace_mcp": {
"command": "uvx",
"args": [
"workspace-mcp"
]
}
}
}Google Workspace MCP Server のツール
120Manually initiate Google OAuth authentication flow. NOTE: This is a legacy OAuth 2.0 tool and is disabled when OAuth 2.1 is enabled. The authentication system automatically handles credential checks and prompts for authentication when needed. Only use this tool if: 1. You need to re-authenticate with different credentials 2. You want to proactively authenticate before using other tools 3. The automatic authentication flow failed and you need to retry In most cases, simply try calling the Google Workspace tool you need - it will automatically handle authentication if required.
Searches messages in a user's Gmail account based on a query. Returns both Message IDs and Thread IDs for each found message, along with Gmail web interface links for manual verification. Supports pagination via page_token parameter.
Retrieves the full content (subject, sender, recipients, body) of a specific Gmail message. Bodies are returned inline and truncated at 20,000 characters. Set full=True to get the complete, untruncated message instead: it is exported to storage and the response carries a short-lived download URL (HTTP transport) or file path (stdio transport) rather than the body, so large messages never stream through the model context. Stateless deployments without GCS staging have no storage, so there full=True returns the untruncated body inline; with GCS staging configured it returns a signed download URL like any other HTTP deployment.
Retrieves the content of multiple Gmail messages in a single batch request. Supports up to 25 messages per batch to prevent SSL connection exhaustion.
Downloads an email attachment and saves it to local disk. In stdio mode, returns the local file path for direct access. In HTTP mode, returns a temporary download URL (valid for 1 hour). May re-fetch message metadata to resolve filename and MIME type.
Sends an email using the user's Gmail account. Supports new emails, replies, and forwards, with optional attachments. Supports Gmail's "Send As" feature to send from configured alias addresses. To forward an existing message, pass forward_message_id. The original subject, body (quoted with a "Forwarded message" header), and attachments are carried over. In forward mode, body (if any) is prepended as a note and subject is optional. Threading, reply, and signature options do not apply when forwarding. THIS TOOL SENDS IMMEDIATELY AND CANNOT SCHEDULE. Gmail's REST API exposes no send-time parameter; Schedule send is a web-UI feature with no API equivalent, so no argument to this tool can defer delivery. Never tell a user a message was scheduled. For "prepare now, deliver later", create a draft with draft_gmail_message. An external scheduler must retain the message data to create and send a new message via send_gmail_message at the chosen time, or call users.drafts.send with the draft ID returned by draft_gmail_message. Alternatively, let the user schedule that draft in the Gmail UI.
Creates a draft email in the user's Gmail account. Supports new, reply, and forward drafts with optional attachments. Supports Gmail's "Send As" feature to draft from configured alias addresses. To save a forward as a draft for review, pass forward_message_id. The original subject, body (quoted with a "Forwarded message" header), and attachments are carried over. In forward mode, body (if any) is prepended as a note, subject is optional, and threading, reply, and signature options do not apply. SCHEDULED SEND IS NOT AVAILABLE. Gmail's REST API exposes no send-time parameter; the Schedule send feature is web-UI only, and a message cannot be placed in the Scheduled folder through the API. Do not claim a message was scheduled. To deliver at a chosen time, create a draft with draft_gmail_message. An external scheduler must retain the message data to create and send a new message via send_gmail_message then, or call users.drafts.send with the draft ID returned by draft_gmail_message. Alternatively, let the user schedule the draft in the Gmail UI.
Retrieves the complete content of a Gmail conversation thread, including all messages. Optionally also returns structured ownership analysis so a caller can determine who sent the last message and who owes whom a response without re-parsing the formatted string or making a second tool call.
Retrieves the content of multiple Gmail threads in a single batch request. Supports up to 25 threads per batch to prevent SSL connection exhaustion.
Lists labels in the user's Gmail account.
Manages Gmail labels: create, update, or delete labels.
Lists all Gmail filters configured in the user's mailbox.
Manages Gmail filters: create, delete, update, and apply to existing mail. - update: Gmail has no filter update API, so the filter is recreated (new one first, then the old one is deleted) and its ID changes. A passed criteria or filter_action replaces that whole object; omit one to keep the old filter's. - apply: runs a filter's label actions on mail ALREADY in the mailbox (Gmail filters only act on new mail). Only matching messages change, not whole conversations as in the web UI. Use filter_id, or criteria + filter_action for an ad-hoc run. Forwarding is never applied. Use dry_run=true first to see the search query and the match count. Needs the gmail.modify scope in addition to gmail.settings.basic.
Adds or removes labels from a Gmail message. To archive an email, remove the INBOX label. To delete an email, add the TRASH label.
Adds or removes labels from multiple Gmail messages in a single batch request, or from whole conversations via thread_ids. message_ids takes MESSAGE ids. Gmail's batch endpoint returns no per-message result and silently ignores ids it does not recognise, so by default this reads the messages back afterwards and reports which ids actually changed. thread_ids applies the change to EVERY message of each conversation (users.threads.modify), e.g. archive a conversation with remove_label_ids=["INBOX"] or mark it read with remove_label_ids=["UNREAD"]. Each thread reports its own outcome. message_ids and thread_ids can be combined; at least one is required.
Searches for files and folders within a user's Google Drive, including shared drives.
Retrieves the content of a specific Google Drive file by ID, supporting files in shared drives. • Native Google Docs, Sheets, Slides → exported as text / CSV. • Office files (.docx, .xlsx, .pptx) → unzipped & parsed with std-lib to extract readable text. • PDFs → text extracted with pypdf when possible; scanned/image-only PDFs fall back to a download hint. • Images → returned as base64 with MIME metadata for multimodal clients. • Any other file → downloaded; tries UTF-8 decode, else notes binary.
Downloads a Google Drive file and saves it to local disk. In stdio mode, returns the local file path for direct access. In HTTP mode, returns a temporary download URL (valid for 1 hour). In stateless mode (no file storage), returns the file itself as an embedded resource, up to WORKSPACE_MCP_STATELESS_INLINE_MAX_BYTES (default 10 MiB, and never more than WORKSPACE_MCP_MAX_FILE_BYTES when that is set). For Google native files (Docs, Sheets, Slides), exports to a useful format: - Google Docs -> PDF (default) or DOCX if export_format='docx' - Google Sheets -> XLSX (default), PDF if export_format='pdf', or CSV if export_format='csv' - Google Slides -> PDF (default) or PPTX if export_format='pptx' For other files, downloads the original file format.
Lists files/folders or shared drive containers, supporting shared drives. If `drive_id` is specified, lists items within that shared drive. `folder_id` is then relative to that drive (or use drive_id as folder_id for root). If `drive_id` is not specified, lists items from user's "My Drive" and accessible shared drives (if `include_items_from_all_drives` is True). Set `resource_type` to "shared_drives" to list shared drive containers instead of folder contents.
Creates a new folder in Google Drive, supporting creation within shared drives.
Creates a new file in Google Drive, supporting creation within shared drives. Accepts direct text content, inline base64 bytes, or a fileUrl to fetch content from. Where local file access is disabled it can instead return an upload URL for the caller to PUT file bytes to. This stores the supplied bytes without converting them to Google Docs, Sheets, or Slides. Use the matching import_to_google_* tool for Google-native conversion.
Imports a file (Markdown, DOCX, TXT, HTML, RTF, ODT) into Google Docs format with automatic conversion. Google Drive automatically converts the source file to native Google Docs format, preserving formatting like headings, lists, bold, italic, etc. Binary sources may be passed directly as base64_content. On a local server a file on disk can also be passed by path (preferred for batch operations, so callers do not need to load full file contents into their context).
Imports a presentation (PPTX, PPT, ODP) into Google Slides format with automatic conversion. Google Drive automatically converts the source presentation to native Google Slides format, preserving slides, layouts, text, and images. Binary sources may be passed directly as base64_content. On a local server a file on disk can also be passed by path (preferred for batch operations, so callers do not need to load full file contents into their context).
Imports a spreadsheet (XLSX, XLS, ODS, CSV, TSV) into Google Sheets format with automatic conversion. Google Drive automatically converts the source spreadsheet to native Google Sheets format, preserving rows, columns, sheets, and values. Binary sources may be passed directly as base64_content. On a local server a file on disk can also be passed by path (preferred for batch operations, so callers do not need to load full file contents into their context).
Gets detailed metadata about a Google Drive file including sharing permissions, parent folder IDs, ownership, and lifecycle timestamps.
Searches for a file by name and checks if it has public link sharing enabled.
Updates metadata, properties, and/or content of a Google Drive file. Providing new content (inline as ``content``, fetched from ``file_url``, or on a local server read from a file path) replaces the file's content in place, preserving the existing file ID, sharing, comments, and links. For native Google Docs/Sheets/Slides the source is uploaded with its source MIME type so the Drive API applies the same format conversion as import_to_google_doc (markdown headings, tables, bold, etc.). For any other file (.md, .txt, .pdf, ...) there is nothing to convert, so the bytes are written back as-is under the file's own MIME type. Metadata and content can be updated in a single call. ``mode='append'``/``'prepend'`` splice ``content`` onto the file's existing text server-side, so only the new text has to be supplied — no need to send the whole file back to rewrite it. Drive shortcuts are handled according to the kind of update: supported resource-local metadata changes (rename, move, trash, star, description, and custom properties) apply to the supplied shortcut, while content replacement follows the shortcut and updates its target. To avoid applying metadata to the wrong resource, a shortcut call cannot combine content with resource-local metadata. Update the shortcut metadata and target content in separate calls.
Gets the shareable link for a Google Drive file or folder.
Consolidated tool for managing Google Drive file and folder access permissions. Supports granting, batch-granting, updating, revoking permissions, and transferring file ownership -- all through a single entry point.
Creates a copy of an existing Google Drive file. This tool copies the template document to a new location with an optional new name. The copy maintains all formatting and content from the original file.
Sets file-level sharing settings and controls link sharing for a Google Drive file or folder. This is a high-level tool for the most common permission changes. Use this to toggle "anyone with the link" access or configure file-level sharing behavior. For managing individual user/group permissions, use share_drive_file or update_drive_permission instead.
Retrieves a list of calendars accessible to the authenticated user.
Retrieves events from a specified Google Calendar. Can retrieve a single event by ID or multiple events within a time range. You can also search for events by keyword by supplying the optional "query" param.
Manages calendar events. Supports creating, updating, deleting, and RSVP.
Manages Out of Office events on Google Calendar. These special events auto-decline meeting invitations and set the user's status to "Out of office" across Google Workspace.
Manages Focus Time events on Google Calendar. These special events auto-decline meeting invitations and, by default, set the user's chat status to Do Not Disturb, helping protect blocks of uninterrupted work time.
Returns free/busy information for a set of calendars.
Creates a new secondary Google Calendar.
Searches for Google Docs by name using Drive API (mimeType filter).
Retrieves content of a Google Doc or a Drive file (like .docx) identified by document_id. - Native Google Docs: Fetches content via Docs API. - Office files (.docx, etc.) stored in Drive: Downloads via Drive API and extracts text. By default, native Google Docs text is index-aligned with the document: empty paragraphs are preserved and every non-text element that occupies an index (inline object, page break, footnote reference, ...) is rendered as one U+FFFC placeholder per index. The document body starts at index 1, so an offset n into the text that follows "--- CONTENT ---" is document index n + 1, and that index can be passed straight to format_text or delete_text. Tables and multi-tab documents interleave separators, so alignment holds up to the first table or tab header. Set preserve_context=True for readable link destinations, internal targets, smart-chip values, table boundaries, headers, footers, footnotes, and object context exposed by the Docs API. This output is plain text, not Markdown, and its offsets must not be used as document editing indices. Comments, revision history, exact visual layout, and chip details hidden by the API are not included. Use get_doc_as_markdown for formatting or comments.
Lists Google Docs within a specific Drive folder.
Creates a new Google Doc and optionally inserts initial content. `content` is inserted as plain text (Markdown is not rendered); use import_to_google_doc to create a doc from Markdown. After creation, the document body starts at index 1. A new empty doc has total length 2 (one section break at index 0, one newline at index 1). To build a rich document after creation, use batch_update_doc with insert_text operations using end_of_segment=true to append content sequentially without calculating indices. Then call inspect_doc_structure to get exact positions before applying formatting in a separate batch call.
Modifies text in a Google Doc - can insert/replace text and/or apply formatting in a single operation. TIP: To append text to the end of the document without calculating indices, set end_of_segment=true. This avoids index calculation errors. For ordinary header/footer text, prefer update_doc_headers_footers. Only pass segment_id when you already have a real header/footer/footnote segment ID from inspect_doc_structure output. Do not guess IDs such as "kix.header" or "kix.footer".
Finds and replaces text throughout a Google Doc. No index calculation required. This is the safest way to update specific text in a document because it does not require knowing any indices. Use this tool when you need to: - Replace placeholder text (e.g., {{TITLE}}) with real content - Update specific words or phrases throughout the document - Make targeted text changes without risk of index errors For building documents from scratch, consider inserting text with unique placeholders via batch_update_doc, then using this tool to replace them.
Inserts structural elements like tables, lists, or page breaks into a Google Doc.
Inserts an image into a Google Doc from Drive or a URL.
Safely creates or updates header/footer text in a Google Doc. This is the default tool for header/footer content. Do NOT use batch_update_doc with create_header_footer just to set header/footer text; that low-level operation is only for advanced section-break workflows and can fail when the default header/footer already exists. This tool handles both creation and update in one call: - If the header/footer does not exist, it is automatically created first. - If the header/footer already exists, its content is replaced. You do NOT need to create a header/footer separately before calling this tool. Simply call it with the desired content and it will work whether the header/footer exists or not.
Executes multiple low-level document operations in a single atomic batch update. For normal header/footer text, prefer update_doc_headers_footers. Only use create_header_footer here for advanced section-break layouts. RECOMMENDED WORKFLOW FOR BUILDING DOCUMENTS: ============================================= To avoid index calculation errors, build documents in phases: PHASE 1 - INSERT ALL CONTENT (use end_of_segment=true, no index math): Append text, section breaks, and page breaks sequentially. Each operation appends to the end of the body. No index needed. Example batch: [ {"type": "insert_text", "end_of_segment": true, "text": "Report Title\n"}, {"type": "insert_text", "end_of_segment": true, "text": "\nExecutive Summary\n"}, {"type": "insert_text", "end_of_segment": true, "text": "Revenue grew 15%.\n"}, {"type": "insert_section_break", "end_of_segment": true, "section_type": "NEXT_PAGE"}, {"type": "insert_text", "end_of_segment": true, "text": "Detailed Analysis\n"} ] PHASE 2 - CREATE HEADERS/FOOTERS (if needed): For normal header/footer text, use update_doc_headers_footers (it auto-creates if missing and writes the content for you). Only include create_header_footer operations in a batch when you are intentionally managing advanced section-break-specific layouts.
Essential tool for finding safe insertion points and understanding document structure. USE THIS FOR: - Finding the correct index for table insertion - Understanding document layout before making changes - Locating existing tables and their positions - Getting document statistics and complexity info - Inspecting structure of specific tabs CRITICAL FOR TABLE OPERATIONS: ALWAYS call this BEFORE creating tables to get a safe insertion index. WHAT THE OUTPUT SHOWS: - total_elements: Number of document elements - total_length: Maximum safe index for insertion - tables: Number of existing tables - table_details: Position and dimensions of each table - headers / footers: Real segment IDs and previews for header/footer editing - tabs: List of available tabs in the document (if no tab_id specified) - empty_paragraphs: newline-only count, excluding existing/suggested object anchors - empty_paragraph_ranges: start/end extents, capped at 100 - empty_paragraph_ranges_truncated: whether ranges were omitted - last_paragraph: is_list_item and is_empty, or null if no body paragraphs Paragraph statistics cover top-level body paragraphs and appear at the top level in basic mode or under "statistics" in detailed mode. Ranges can include required empty paragraphs. Before cleanup, use detailed=true to check adjacent elements, formatting, and object anchors. Preserve the final newline and newlines before tables, tables of contents, or section breaks. If the body ends with an empty paragraph, that paragraph's range has end == total_length; truncation may omit that range. WORKFLOW FOR TABLE INSERTION: Step 1: Call this function Step 2: Note the "total_length" value Step 3: Use an index < total_length for table insertion Step 4: Create your table FORMATTING WORKFLOW: After inserting all text via batch_update_doc with end_of_segment=true, call this tool with detailed=true to get exact start_index and end_index for every paragraph. Use those indices directly in format_text and update_paragraph_style operations in a second batch_update_doc call. HEADER/FOOTER WORKFLOW: For ordinary header/footer text, use update_doc_headers_footers. If you need low-level segment editing, call this tool first and use the real segment_id values returned under headers/footers. Do not invent IDs. The detailed output includes elements[].start_index and elements[].end_index with text_preview for each paragraph, making it easy to identify which ranges to format. SUB-PARAGRAPH FORMATTING: text_preview is truncated to 100 characters by default. To compute the index of a token inside a longer paragraph, pass preview_chars=0 or None for untruncated text, then add the token's UTF-16 offset within that text to the paragraph's start_index (non-BMP characters such as emoji occupy two UTF-16 units).
Return runtime/source information for diagnosing stale MCP server instances. This is a temporary diagnostic tool intended to verify which code checkout the running MCP server has loaded.
Creates a table and populates it with data in one reliable operation. CRITICAL: YOU MUST CALL inspect_doc_structure FIRST TO GET THE INDEX! MANDATORY WORKFLOW - DO THESE STEPS IN ORDER: Step 1: ALWAYS call inspect_doc_structure first Step 2: Use the 'total_length' value from inspect_doc_structure as your index Step 3: Format data as 2D list: [["col1", "col2"], ["row1col1", "row1col2"]] Step 4: Call this function with the correct index and data EXAMPLE DATA FORMAT: table_data = [ ["Header1", "Header2", "Header3"], # Row 0 - headers ["Data1", "Data2", "Data3"], # Row 1 - first data row ["Data4", "Data5", "Data6"] # Row 2 - second data row ] CRITICAL INDEX REQUIREMENTS: - NEVER use index values like 1, 2, 10 without calling inspect_doc_structure first - ALWAYS get index from inspect_doc_structure 'total_length' field - Index must be a valid insertion point in the document DATA FORMAT REQUIREMENTS: - Must be 2D list of strings only - Each inner list = one table row - All rows MUST have same number of columns - Use empty strings "" for empty cells, never None - Use debug_table_structure after creation to verify results
ESSENTIAL DEBUGGING TOOL - Use this whenever tables don't work as expected. USE THIS IMMEDIATELY WHEN: - Table population put data in wrong cells - You get "table not found" errors - Data appears concatenated in first cell - Need to understand existing table structure - Planning to use populate_existing_table WHAT THIS SHOWS YOU: - Exact table dimensions (rows × columns) - Each cell's position coordinates (row,col) - Current content in each cell - Insertion indices for each cell - Table boundaries and ranges HOW TO READ THE OUTPUT: - "dimensions": "2x3" = 2 rows, 3 columns - "position": "(0,0)" = first row, first column - "current_content": What's actually in each cell right now - "insertion_index": Where new text would be inserted in that cell WORKFLOW INTEGRATION: 1. After creating table → Use this to verify structure 2. Before populating → Use this to plan your data format 3. After population fails → Use this to see what went wrong 4. When debugging → Compare your data array to actual table structure
Exports a Google Doc to PDF format and saves it to Google Drive.
Apply paragraph-level formatting, heading styles, and/or list formatting to a range in a Google Doc. This tool can apply named heading styles (H1-H6) for semantic document structure, create bulleted or numbered lists with nested indentation, and customize paragraph properties like alignment, spacing, and indentation. All operations can be applied in a single call.
Reads a Google Doc and returns it as clean Markdown with optional comment context. Unlike get_doc_content which returns plain text, this tool preserves document formatting as Markdown: headings, bold/italic/strikethrough, links, code spans, ordered/unordered lists with nesting, and tables. When comments are included (the default), each comment's anchor text — the specific text the comment was attached to — is preserved, giving full context for the discussion.
Manage document tabs: create, rename, delete, or populate from Markdown.
List all comments from a Google Document (optional max_comments to limit results).
Manage comments on a Google Document.
Lists spreadsheets from Google Drive that the user has access to.
Gets information about a specific spreadsheet including its sheets.
Reads values from a specific range in a Google Sheet.
Modifies values in a specific range of a Google Sheet - can write, update, or clear values, or insert Smart Chips (Drive files/folders or People).
Applies formatting to a range: colors, number formats, text wrapping, alignment, text styling, cell merging/unmerging, borders, and dropdown lists. Colors accept hex strings (#RRGGBB). Number formats follow Sheets types (e.g., NUMBER, CURRENCY, DATE, PERCENT). If no sheet name is provided, the first sheet is used.
Manages conditional formatting rules on a Google Sheet. Supports adding, updating, and deleting conditional formatting rules via a single tool.
Creates a new Google Spreadsheet.
Creates a new sheet or duplicates an existing sheet (user_google_email: str, spreadsheet_id: str, sheet_name: Optional[str] = None, source_sheet_name: Optional[str] = None, insert_sheet_index: Optional[int] = None).
Lists all structured tables in a spreadsheet with their IDs, names, ranges, and column details. Use this to find table IDs for append_table_rows.
Appends rows to a structured table in a Google Sheet. The rows are added to the end of the table body, automatically extending the table range. Use list_sheet_tables first to find the table ID.
Manages an existing sheet tab: rename, delete, hide, unhide, reorder, or style (tab color and gridlines). Use create_sheet to add a tab, and resize_sheet_dimensions for row and column level changes. This tool operates on the tab itself.
Manages sheet-level dimension properties: resize columns/rows, auto-resize to fit content, freeze rows/columns, hide/unhide rows/columns, and insert/delete rows/columns.
Moves rows from one sheet to another within the same spreadsheet. The move is performed in a single batchUpdate (copyPaste followed by deleteDimension). Note: batchUpdate executes requests sequentially but does not roll back on partial failure — if the copy succeeds but the delete fails, rows may be duplicated. Formulas, data types, and formatting are preserved (unlike a values.get/append round-trip). Row numbers are 1-based (matching the spreadsheet UI).
Manages the lifecycle of named ranges in a Google Sheet: list, create, update, or delete.
List all comments from a Google Spreadsheet (optional max_comments to limit results).
Manage comments on a Google Spreadsheet.
Lists Google Chat spaces (rooms and direct messages) accessible to the user.
Retrieves messages from a Google Chat space.
Sends a message to a Google Chat space, or edits a message already sent there.
Searches for messages in Google Chat spaces by text content and/or time range.
Adds an emoji reaction to a Google Chat message.
Downloads an attachment from a Google Chat message and saves it to local disk. In stdio mode, returns the local file path for direct access. In HTTP mode, returns a temporary download URL (valid for 1 hour).
Create a new form using the title given in the provided form message in the request.
Get a form.
Updates the publish settings of a form.
Get one response from the form.
List a form's responses.
Apply batch updates to a Google Form. Supports adding, updating, and deleting form items, as well as updating form metadata and settings. This is the primary method for modifying form content after creation.
Create a new Google Slides presentation.
Get details about a Google Slides presentation.
Apply batch updates to a Google Slides presentation.
Get details about a specific page (slide) in a presentation.
Generate a thumbnail URL for a specific page (slide) in a presentation.
List all comments from a Google Presentation (optional max_comments to limit results).
Manage comments on a Google Presentation.
List all task lists for the user.
Get details of a specific task list.
Manage task lists: create, update, delete, or clear completed tasks.
List all tasks in a specific task list.
Get details of a specific task.
Manage tasks: create, update, delete, or move tasks within task lists.
List contacts for the authenticated user.
Get detailed information about a specific contact.
Search contacts by name, email, phone number, or other fields.
Create, update, or delete a contact. Consolidated tool replacing create_contact, update_contact, and delete_contact.
List contact groups (labels) for the user.
Get details of a specific contact group including its members.
Batch create, update, or delete contacts. Consolidated tool replacing batch_create_contacts, batch_update_contacts, and batch_delete_contacts.
Create, update, delete a contact group, or modify its members. Consolidated tool replacing create_contact_group, update_contact_group, delete_contact_group, and modify_contact_group_members.
Performs a search using Google Custom Search JSON API.
Retrieves metadata about a Programmable Search Engine.
Read Apps Script projects and their source files.
Update the source files of an Apps Script project.
Executes a function in a deployed script.
Create, update, or delete Apps Script deployments.
List deployments for an Apps Script project.
Create or delete an Apps Script project.
Create immutable version snapshots of a script project. Versions capture a snapshot of the current script code; once created they cannot be modified.
List or retrieve immutable version snapshots of a script project.
Read execution activity for scripts: recent processes or aggregate metrics.
Generates Apps Script code for creating triggers. The Apps Script API cannot create triggers directly - they must be created from within Apps Script itself. This tool generates the code you need. To list or remove existing triggers without opening the editor, use `manage_script_trigger` (action="list" / action="delete") instead.
List or delete the current user's installable triggers on a script project. The Apps Script REST API has no triggers resource, so in dev_mode both actions provision (or refresh) a small helper file and run it via the Execution API - the only way to inspect or change trigger state without opening the editor. Neither action is read-only: the helper file may be written into the project on first use. To create a trigger, use `generate_trigger_code` to produce the setup code, add it with `manage_script_content`, then run it with `run_script_function`.
概要
Google Workspace MCP connects AI assistants, coding agents, and developer tools to Gmail, Google Drive, Calendar, Docs, Sheets, Slides, Forms, Chat, Apps Script, Tasks, Contacts, and Google Custom Search through one MCP server.
It provides deep read and write access across all 12 service groups, including fine-grained document editing, email and calendar management, file sharing, spreadsheet formatting, presentation creation, form responses, Chat spaces, Apps Script execution, and cross-service workflows. The same toolset is also available through a full-featured CLI for coding agents such as Claude Code and Codex.
Ready to connect your first client? Follow the interactive Quick Start Guide, or explore every tool and configuration option in the complete documentation.
What is Google Workspace MCP Server?
Google Workspace MCP Server is an open-source Model Context Protocol server that gives MCP-compatible clients a unified interface to Google Workspace. Instead of configuring a separate integration for every Google service, you can run one server and let an AI assistant securely search, read, create, edit, organize, and share Workspace content on behalf of the signed-in user.
The server works with free Google accounts and Google Workspace plans. It supports local use, direct CLI automation, and centrally hosted multi-user deployments with OAuth 2.1.
How to use Google Workspace MCP Server
The fastest path is the five-step Workspace MCP Quick Start, which walks through installation, Google OAuth setup, server startup, client connection, and validation.
You will need Python 3.10 or newer, uv/uvx, and a Google Cloud project with OAuth credentials. For a confidential OAuth client, a basic modern HTTP setup looks like this:
export GOOGLE_OAUTH_CLIENT_ID="your-client-id"
export GOOGLE_OAUTH_CLIENT_SECRET="your-client-secret"
export MCP_ENABLE_OAUTH21=true
uvx workspace-mcp --transport streamable-http --tool-tier core
Streamable HTTP is the recommended transport for modern MCP clients, OAuth 2.1, and multi-user deployments. Claude Desktop, Claude on the web, and Claude Cowork can connect through a custom Connector; Claude Code, VS Code, MCP Inspector, and other compatible clients can connect directly to the HTTP endpoint. Stdio remains available as a legacy fallback.
Start with the core tool tier, move to extended or complete when you need more operations, or load only selected services with --tools. Public PKCE clients, reverse proxies, and hosted deployments require additional authentication settings, so use the Quick Start Guide for setup and the authentication documentation for production configuration.
Key features of Google Workspace MCP Server
- 12 service groups in one server: Gmail, Drive, Calendar, Docs, Sheets, Slides, Forms, Chat, Apps Script, Tasks, Contacts, and Custom Search
- Deep Workspace operations: search, read, create, edit, format, comment, share, organize, send, publish, and automate
- Full CLI support: list and invoke tools directly from terminals, scripts, Claude Code, Codex, and other coding-agent workflows
- Modern authentication: OAuth 2.0, OAuth 2.1 with PKCE, automatic token refresh, multi-user bearer authentication, external OAuth providers, and service-account domain-wide delegation
- Flexible tool control: cumulative
core,extended, andcompletetiers, per-service selection, read-only mode, and granular permission levels - Deployment-ready: streamable HTTP, stateless containers, reverse proxies, encrypted OAuth storage backends, and centrally hosted multi-user instances
- Private by default: uses your OAuth client and Google Cloud project; the default data path is directly from your infrastructure to Google APIs
- Open source: MIT licensed, with no license server, SaaS dependency, usage reporting, or analytics
See the complete Workspace MCP documentation for tool parameters, permission controls, authentication modes, environment variables, and deployment guidance.
Google Workspace MCP use cases
- Search Gmail for a conversation, save its attachments to Drive, and summarize the result in a Google Doc
- Find open Calendar time, create an event, add attendees, and attach relevant Drive files
- Create and format Docs, Sheets, Slides, and Forms from an AI assistant or coding agent
- Draft, send, label, and organize email while limiting the server to only the Gmail permissions you approve
- Manage Drive files, folders, shared drives, permissions, imports, exports, and shareable links
- Build Apps Script projects and execute cross-application Workspace automations
- Create internal assistants that work across company email, documents, tasks, contacts, calendars, and Chat spaces
- Run read-only research and reporting workflows with restricted OAuth scopes
- Host one secure OAuth 2.1 MCP endpoint for a team or an entire organization
- Use the CLI in scripts and coding-agent workflows without operating through a chat interface
Google Workspace MCP Server FAQ
What are the runtime requirements?
Google Workspace MCP requires Python 3.10 or newer, uv/uvx, and a Google Cloud project with OAuth credentials. You must enable the Google APIs for the services you plan to use. The Quick Start Guide walks through the complete setup.
Does the server send Workspace data to a third party?
By default, no. The server uses your own OAuth client and sends requests directly from your infrastructure to Google APIs on behalf of the authenticated user. There is no built-in analytics, usage reporting, license server, token relay, or required SaaS backend. Optional OpenTelemetry tracing remains disabled unless you install it and explicitly configure an OTLP endpoint.
How does authentication work?
You provide your own Google OAuth client. The server supports OAuth 2.0 for legacy local flows and OAuth 2.1 with PKCE and bearer tokens for modern or multi-user deployments. It also supports public PKCE clients without a Google client secret, external OAuth 2.1 providers, and service accounts with domain-wide delegation. Credential and OAuth-state storage can be configured for local disk, memory, GCS, or Valkey/Redis depending on the deployment.
Can I restrict what an AI assistant can access?
Yes. Use --read-only to request read-only scopes and remove write tools, select individual services with --tools, or apply per-service permission levels such as gmail:send or drive:readonly. Tool tiers provide another way to expose only the operations you need.
Can I host one server for my organization?
Yes. OAuth 2.1 multi-user authentication, stateless mode, external OAuth provider support, reverse-proxy configuration, and pluggable encrypted storage backends are designed for centralized deployments. You retain control of the infrastructure, OAuth client, scopes, credentials, and network path.
What transports are supported?
The server supports streamable HTTP and stdio. Streamable HTTP is recommended for modern clients and is required for OAuth 2.1. Stdio is retained as a legacy fallback for older or incomplete MCP clients.
Which Google accounts are supported?
The server supports free Google accounts and Google Workspace editions, including Starter, Standard, Plus, Enterprise, and Nonprofit plans. Some Google services have account-specific requirements; for example, Google Chat tools require a Workspace account and additional Chat app configuration.
Where can I find every available tool and parameter?
Browse the searchable Google Workspace MCP documentation for the full tool reference, examples, authentication settings, transport options, CLI usage, permissions, and deployment guides.
Get started with Google Workspace MCP
Connect your first AI client in five steps with the interactive Quick Start Guide. For advanced setup, production hosting, and the complete API reference, visit the Workspace MCP documentation.
「その他」の他のコンテンツ
Figs, Relationship Expert
Figs O'Sullivan, LMFTCouples therapist Figs O'Sullivan (LMFT, 17 years, endorsed by Sue Johnson, creator of EFT) inside your AI. Ask relationship questions answered from his clinical writing, name the pattern in a fight, get a three-step rep
Senso MCP for Joomla
sensomediaSenso MCP turns a Joomla 5 or 6 site into a remote MCP server. It runs inside Joomla itself (no extra server, no Node.js bridge) and is secured with OAuth 2.1 and Joomla's own ACL. The free pack lets an AI assistant crea
Health by Dr Brad
Dr Brad StanfieldHealth by Dr Brad works out a health plan from your own results: which screening is due at your age, and evidence-based suggestions, each with the reason and the published guideline or study behind it. Your record is one
Maestro
mobile-dev-incPainless E2E Automation for Mobile and Web
MCP Toolbox for Databases
googleapisMCP Toolbox for Databases is an open source MCP server for databases.
コメント