Skip to main content
Glama
better-workspace-mcp

better-workspace-mcp

by n24q02m

ARCHIVED 2026-09-13 — This repository is no longer maintained. Use the official Google Workspace APIs instead of this MCP server. Existing installations keep working but receive no updates or support.

Project

Tagline

Tag

better-semantic-release

Drop-in python-semantic-release fork with built-in release-safety guards (orp...

Tooling

claude-plugins

Claude Code plugin marketplace for the n24q02m MCP servers -- install web sea...

Marketplace

crg

Knowledge graph for token-efficient code reviews -- semantic search and call-...

MCP

email

IMAP/SMTP email for AI agents -- read, send, organize folders, and manage att...

MCP

fastretrieval

Fast multi-model retrieval runtime: ONNX and GGUF embeddings, reranking, and ...

Tooling

hull

hull is the shared monorepo core for the wet, crg and mnemo self-hosted stack

MCP

jules-task-archiver

Chrome Extension for bulk operations on Jules tasks via batchexecute API -- a...

Tooling

mnemo

Persistent AI memory with hybrid search. Open, free, unlimited.

MCP

notion

Markdown-first Notion for AI agents -- pages, databases, blocks, and comments...

MCP

skret

Secrets without the server.

CLI

tacet

A self-distilling neuro-symbolic cascade that amortises LLM cost across knowl...

Tooling

telegram

Telegram for AI agents -- messages, chats, media, and contacts across both bo...

MCP

tools-docs

Docs portal for the n24q02m open-source tool stack (skret, MCP plugins, hull,...

Tooling

wet

Open-source MCP server for AI agents: web search, content extraction, and lib...

MCP

workspace

Google Workspace MCP server (Docs/Drive/Calendar/Gmail/Sheets/Slides/Tasks/Ch...

MCP

Table of contents

Related MCP server: google-workspace-mcp

Install

The server runs in two modes: stdio (default, single-user, one Google OAuth client of your own) and HTTP (multi-user, per-user bearer tokens). For stdio, add it to your MCP client config:

{
  "mcpServers": {
    "workspace": {
      "command": "npx",
      "args": ["--yes", "@n24q02m/workspace@latest"],
      "env": {
        "GOOGLE_OAUTH_CLIENT_ID": "<your-client-id>.apps.googleusercontent.com",
        "GOOGLE_OAUTH_CLIENT_SECRET": "<your-client-secret>"
      }
    }
  }
}

The package is @n24q02m/workspace (formerly @n24q02m/better-workspace-mcp; old package receives one final release then freezes). The MCP Registry id is now io.github.n24q02m/workspace; the entry published under io.github.n24q02m/better-workspace-mcp stays on the registry as a frozen historical release, and the legacy better-workspace-mcp command is kept as an alias of workspace.

Those two values come from an OAuth 2.0 client of type Desktop app, created in the Google Cloud Console under APIs & Services → Credentials. Desktop is the right type because the server receives the consent redirect on a loopback address, not on a public URL. Enable the Workspace APIs you plan to call on the same project, and add yourself as a test user while the consent screen is unpublished.

The first run prints a Google consent URL to stderr (your MCP client's server log) and waits; open it in a browser and finish the consent. The refresh token is stored encrypted on your machine under ~/.workspace/ (a store an earlier release wrote under ~/.better-workspace/ is copied forward on the next start, and the older ~/.better-workspace-mcp/ is read as a fallback), so later runs start without asking again.

The same server can also serve several people over HTTP -- see Remote (HTTP mode).

Install with an AI agent -- paste this to your AI coding agent:

Install MCP server workspace following the steps at https://raw.githubusercontent.com/n24q02m/claude-plugins/main/plugins/workspace/setup-with-agent.md

Remote (HTTP mode)

Besides stdio, the server runs as a multi-user HTTP service. Each request carries a bearer token from a users file (WORKSPACE_USERS_FILE), and each user's Google credentials are kept in their own bucket, keyed by that user's namespace, so one deployment serves several people without them sharing an account. Google consent for an HTTP account runs on the host, with workspace auth --user <namespace>.

Point your MCP client at the host you deployed it on:

{
  "mcpServers": {
    "workspace": {
      "type": "http",
      "url": "https://<your-host>/mcp"
    }
  }
}

Modes

Two, and only two. There is no proxy or daemon mode: stdio speaks the MCP stdio transport directly, with no HTTP hop inside it.

Mode

Selected by

Serves

Google OAuth client

stdio

the default

one user, on their own machine

Desktop app

http

--http, MCP_TRANSPORT=http, or TRANSPORT_MODE=http

several users, one credential bucket per user namespace

Web application

Two OAuth clients, not one

Google binds redirect URIs to the type of the OAuth client, so the two modes cannot share one. stdio uses a Desktop client, which may redirect to any loopback port. HTTP uses a Web client. Keep the two credentials under separate names: pointing GOOGLE_OAUTH_CLIENT_ID/_SECRET at the Web client on a machine that also runs stdio breaks every stdio install. That is why docker-compose.http.yml reads the Web pair as GOOGLE_OAUTH_WEB_CLIENT_ID/_SECRET and renames it on the way into the container.

Environment

Variable

Required

What it does

GOOGLE_OAUTH_CLIENT_ID

yes

Web client id. The server starts without it, but workspace auth refuses to add an account and stored tokens cannot be refreshed.

GOOGLE_OAUTH_CLIENT_SECRET

yes

Web client secret. Same.

CREDENTIAL_SECRET

yes

Derives the encryption key of each user's credential store. The server refuses to start without it, rather than fail in the middle of a user's consent.

WORKSPACE_USERS_FILE

yes in HTTP

Users file that maps each bearer token to a user namespace (hull multi auth). Unset uses hull's default location.

WORKSPACE_DATA_DIR

no

Base directory for the credential store (<dir>/.workspace/, one encrypted blob per user). Unset uses the home directory. A relative value is resolved under the home directory and must stay inside it, so use an absolute path for a mounted volume.

PORT

no

Listen port. 0 (the default) asks the OS for a free one.

HOST

no

Bind address. Defaults to loopback, so a container needs 0.0.0.0.

Credentials are stored on local disk under ~/.workspace/, which is the data volume in the container. A store from an earlier release (~/.better-workspace/) is copied forward on startup, and the older ~/.better-workspace-mcp/ is read as a fallback.

docker-compose.http.yml in this repo is these variables written out as a runnable overlay.

Tools

One composite tool per Workspace domain, plus config and help:

Tool

What it covers

docs

Google Docs -- getText, create, writeText, getSuggestions, replaceText, formatText

drive

Files and folders -- search, findFolder, createFolder, moveFile, renameFile, trashFile, downloadFile, getComments

calendar

Events -- listCalendars, listEvents, getEvent, createEvent, updateEvent, deleteEvent, respondToEvent, findFreeTime

gmail

Mail -- search, get, send, createDraft, sendDraft, modify, batchModify, modifyThread, downloadAttachment, listLabels, createLabel

sheets

Spreadsheets, read-only -- getText, getRange, getMetadata

slides

Presentations -- 19 actions covering slides, text, shapes, images, tables, and speaker notes

tasks

Task lists and tasks -- listTaskLists, listTasks, createTask, updateTask, completeTask, deleteTask

chat

Google Chat -- listSpaces, findSpaceByName, setUpSpace, getMessages, listThreads, sendMessage, sendDm, findDmByEmail

people

Profile lookups -- getMe, getUserProfile, getUserRelations

forms

Forms -- create, get, batchUpdate, listResponses, getResponse

time

Local date/time/timezone helpers (no Google account needed)

config

Credential state and account management (CLI: config-tool)

help

Full documentation for any tool

Questions are added to a form with forms(action="batchUpdate"), not at create; responses are read-only, because the Forms API cannot write one. Listing or deleting forms goes through drive.

Response size

A Google read can return megabytes from one call, so every action whose payload can grow is bounded by default:

  • 20,000 characters per response. Raise it with max_chars (1--100,000), or pass full: true for the complete payload. The same envelope docs exposes as limit/preview and sheets as max_chars/full.

  • Listings that page keep their page size (default 20, clamped to 1--100) and always return their nextPageToken; listings with no cursor (getImages, findFolder, findSpaceByName, listCalendars, getUserRelations, free/busy slots, labels) also cap how many items one response carries (50--100).

  • Truncation is never silent. A cut payload carries _truncation: {truncated, total_characters, returned_characters, omitted_items?, note} whose note reads [Content truncated: showing <kept> of <total> characters (<omitted> omitted). ...]. A JSON payload stays parseable, and a pagination cursor is never cut.

workspace gmail get --messageId <id> --max_chars 60000 --json   # a longer body
workspace slides getText --presentationId <id> --full --json     # the whole deck

CLI

Every tool is also a shell subcommand, so the same operations work without an MCP client. workspace <tool> <action> turns those parameters into the tool's input object and calls the same handler the server registers -- one dispatch, two front doors, so credential gating and validation behave identically.

workspace <tool> <action> [--param <value>]... [--json]

--json prints one envelope on stdout -- {"tool", "action", "result"}, where result is the tool's own payload (parsed when the tool emits JSON, the raw text otherwise). Without it you get the tool's text as-is. A tool failure prints to stderr and exits 1; argv the CLI cannot parse exits 2. workspace --help lists the tool subcommands and the operator commands (auth, config, status, logout, doctor, token) separately, and workspace <tool> --help prints that tool's actions and options.

One tool name is already taken. config is the operator command that reads the credential store hull manages (workspace config status); the MCP config tool is a different surface -- credential lifecycle plus the Google accounts the server acts as. It keeps a console path under the alias config-tool, so neither command can silently stand in for the other:

workspace config-tool status --json
workspace config-tool account_list

Group

Example

docs

workspace docs getText --documentId <document-id> --limit 2000

drive

workspace drive search --query "name contains 'Q3 report'" --pageSize 5

calendar

workspace calendar listCalendars --json

chat

workspace chat listSpaces --json

gmail

workspace gmail search --query "is:unread" --maxResults 5 --json

sheets

workspace sheets getRange --spreadsheetId <spreadsheet-id> --range "Sheet1!A1:C10"

slides

workspace slides getMetadata --presentationId <presentation-id>

tasks

workspace tasks listTasks --taskListId <task-list-id> --maxResults 20

people

workspace people getMe --json

forms

workspace forms listResponses --formId <form-id> --pageSize 10

time

workspace time getCurrentTime --json

config-tool

workspace config-tool status --json (the MCP config tool)

help

workspace help gmail

Values are coerced from the tool's own schema: --maxResults 5 is a number, --includeSpamTrash (also --includeSpamTrash true, or --no-includeSpamTrash) is a boolean, and --labelIds INBOX,IMPORTANT (or a JSON array) is a list. Every domain accepts --account <email> to choose the Google account, exactly like the tool's account parameter.

Quick start

Check that the credentials landed, with the config tool. Before the first consent this reports awaiting_setup; afterwards it names the account the server is acting as:

{ "action": "status" }

From a shell the same call is workspace config-tool status --json, and it works before consent -- it reads the store, it does not need Google.

Then call a domain tool. On docs, create returns the new document's ID, which getText reads back:

{ "action": "create", "title": "Notes", "content": "First line." }
{ "action": "getText", "documentId": "<id-from-create>" }

time, config, and help need no Google account, so they answer even before consent -- time is the quickest check that the server is wired up at all:

{ "action": "getCurrentTime" }

To act as a second Google account, see Multi-account.

Multi-account

Every domain tool takes an account parameter -- the email of the Google account the call acts as. Omit it and the call runs against the primary account.

{ "action": "search", "query": "is:unread", "account": "work@example.com" }
{ "action": "search", "query": "is:unread" }

Those two gmail calls read two different mailboxes: the first work@example.com, the second whichever account is primary.

Accounts are managed through the config tool:

Call

Effect

config(action="account_add")

In stdio, returns a URL to open; completing the Google consent there adds one more account. Over HTTP it returns the workspace auth --user <namespace> command to run on the host.

config(action="account_list")

The configured accounts and which one is primary.

config(action="account_remove", account="<email>")

Forget one account.

config(action="account_set_default", account="<email>")

Make one account the primary.

From a shell, the same four calls:

workspace config-tool account_add
workspace config-tool account_list
workspace config-tool account_remove --account <email>
workspace config-tool account_set_default --account <email>

config-tool, not config: the plain name is the operator command that reads the credential store.

The first account authorized becomes the primary. Removing the primary promotes one of the remaining accounts; removing the last one puts the server back to awaiting setup. Naming an account that is not configured is an error that names it -- the call is never rerouted to the primary, because a silent fallback would act on the wrong mailbox.

account_add chooses its flow by transport. In stdio it starts a temporary loopback consent server and returns the URL. Over HTTP it does not serve consent; it returns the workspace auth --user <namespace> command for the host.

value="primary" is stdio only; remote callers change the default with account_set_default.

Coming from an earlier single-account build

Credentials stored by a build from before multi-account support are one flat blob of tokens. The first run afterwards adopts that blob into the multi-account layout under the account's email. The email comes from the stored id_token when it is present -- no network needed, and that is the usual case. Otherwise the server asks Google's userinfo endpoint, which needs network access and a token that is still valid or refreshable.

If neither works, the server reports itself as awaiting setup and starts the consent flow again, printing its URL to stderr, even though the stored token may still be fine. Completing that consent works, and nothing is discarded: the account you just authorized is stored and becomes the primary, while the old tokens are carried across under the key (unidentified) rather than being overwritten. config(action="account_list") then shows two entries:

{ "accounts": ["(unidentified)", "you@example.com"], "primary": "you@example.com" }

Nothing routes to (unidentified): it is never promoted to primary while a real account remains, and no call reaches it unless you name it explicitly. It is there so a credential whose owner could not be determined is not silently thrown away. Remove it once the re-authorized account is working:

{ "action": "account_remove", "account": "(unidentified)" }

After a successful adoption -- the usual case -- config(action="account_list") shows exactly one account, and it is the primary.

The consent screen has requested the Google Forms scopes since before the forms tool existed, so it arrived without a second trip through consent.

Accounts authorized before those scopes were added are covered too, but by a different route: Google accepts https://www.googleapis.com/auth/drive in place of the Forms scopes for every Forms method this server calls, and full drive has been on the consent screen since the first release. Google never widens a token it has already issued, so if a grant covers neither -- a user may withhold individual restricted scopes at the consent screen -- the first forms call returns a 403 (Request had insufficient authentication scopes). Re-authorize just that account with config(action="account_add"), signing in as the same account; the record is replaced in place, so nothing else changes.

Documentation

Full docs at docs.n24q02m.com/servers/workspace/:

  • Setup -- plugin install, Docker stdio, and Docker HTTP methods

Every tool also documents itself at runtime: call help for the full reference on any of them, including the exact parameters each action takes.

Deployment

No hosted endpoint is operated for this server. The supported paths are the stdio install above and a self-hosted HTTP deployment of your own.

To self-host HTTP mode, run the compose overlay from this repository:

docker compose -f docker-compose.yml -f docker-compose.http.yml up

It builds the http image locally and publishes it on host loopback only, at 127.0.0.1:8087. Set GOOGLE_OAUTH_WEB_CLIENT_ID, GOOGLE_OAUTH_WEB_CLIENT_SECRET, CREDENTIAL_SECRET, and WORKSPACE_USERS_FILE through your shell or an .env file (see Environment). The users file maps each bearer token to a user namespace; accounts are added on the host with workspace auth --user <namespace>. For use beyond the local machine, put your own TLS reverse proxy on the same host in front of it.

Each release also publishes multi-arch container images to Docker Hub and GHCR; their default stage is the http server.

Contributing

See CONTRIBUTING.md.

License

Apache-2.0 © n24q02m

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers