Vexa
Meeting bot and transcripts for Google Meet, Teams and Zoom. Live or after, speakers labelled.
Beschreibung vom Entwickler (englisches Original)
67 curated MCP tools for Google Workspace: Gmail, Calendar, Drive, Sheets, Docs, Slides, Tasks.
Diese drei werden nie zu einer Punktzahl zusammengefasst: Meinungen, Messungen und Beliebtheit sind verschiedene Dinge.
Für diesen Server haben wir noch keine Zusammenfassung in einfacher Sprache geschrieben. Unten stehen die Fakten, die wir aus der Registry gelesen haben.
Wähle die KI-App, die du nutzt, und kopiere die Einstellungen.
Du brauchst
Welche App nutzt du?
Wähle oben deine App. Wir merken sie uns, damit der Button beim nächsten Mal „In … installieren“ anzeigt.
Öffne Claude Desktop → Settings → Developer → Edit Config und füge dies in claude_desktop_config.json ein
{
"mcpServers": {
"gws": {
"command": "npx",
"args": [
"-y",
"gws-mcp-server",
"--services",
"drive,sheets,calendar,docs,slides,gmail,tasks"
]
}
}
}Führe das in deinem Terminal aus
claude mcp add gws -- npx -y gws-mcp-server --services 'drive,sheets,calendar,docs,slides,gmail,tasks'In .cursor/mcp.json in deinem Projekt oder in ~/.cursor/mcp.json einfügen
{
"mcpServers": {
"gws": {
"command": "npx",
"args": [
"-y",
"gws-mcp-server",
"--services",
"drive,sheets,calendar,docs,slides,gmail,tasks"
]
}
}
}Dein Browser fragt, ob er die App öffnen darf, dann zeigt die App die Einstellungen und bittet dich um Bestätigung. Schlüsselwerte bleiben leer, damit du sie selbst einträgst.
In .vscode/mcp.json in deinem Projekt einfügen
{
"servers": {
"gws": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"gws-mcp-server",
"--services",
"drive,sheets,calendar,docs,slides,gmail,tasks"
]
}
}
}Dein Browser fragt, ob er die App öffnen darf, dann zeigt die App die Einstellungen und bittet dich um Bestätigung. Schlüsselwerte bleiben leer, damit du sie selbst einträgst.
Dieser Server läuft auf deinem Computer, und ChatGPT kann sich nur mit Servern im Internet verbinden. Wähle eine andere App.
Deine Auswahl bleibt in einem Cookie auf diesem Gerät; kein Konto nötig.
Diese Einstellungen werden aus dem Registry-Eintrag des Entwicklers erzeugt. Wenn in der README etwas anderes steht, folge der README.
Läuft auf deinem Computer; die Tool-Liste erscheint nach der Installation in deiner KI-App.
Wir prüfen Informationen, die jeder sehen kann; den Code haben wir nicht gelesen. Checks, die nicht zutreffen, werden nicht mitgezählt. · Geprüft am 7. Nov. 2026
Der Registry-Name wurde mit dem GitHub-Konto conorbronsdon bestätigt, dem das Repo gehört
Dieses Repo haben wir noch nicht von GitHub gelesen
Dieses Repo haben wir noch nicht von GitHub gelesen
Dieser Server läuft auf deinem Computer; es gibt keinen Endpunkt im Internet
Läuft auf deinem Computer und nutzt stattdessen lokale Einstellungen
Läuft auf deinem Computer; Tool-Listen können wir nur von Endpunkten im Internet lesen
Verglichen mit allen aktiven Servern dieser Kategorie in der Registry. Sterne zählen nur, wenn dem Herausgeber das Repo gehört.
Er hat 1 von 1 zutreffenden Basischecks bestanden. Den Code haben wir nicht gelesen. Bevor du einen Schlüssel oder Token hinzufügst, sieh nach, welche Zugriffe der Server verlangt, und nutze einen Schlüssel, der nur die Berechtigungen hat, die du brauchst.
Die Registry nennt keinen erforderlichen Schlüssel, manche Server bitten dich aber bei der ersten Nutzung, dich anzumelden.
Er läuft auf deinem Computer. Deine KI-App startet ihn bei Bedarf, und die Daten laufen nicht über die Server des Entwicklers (außer der Server ruft selbst einen Onlinedienst auf).
Das GitHub-Repo wurde zuletzt – geändert, und die neueste Version (0.4.1) kam vor 44 Tagen in die Registry.
Apps, die MCP unterstützen, etwa Claude Desktop, Claude Code, Cursor und VS Code. Außerdem brauchst du Node.js auf deinem Computer.
In derselben Kategorie: Vexa, emailmd, MCP WhatsApp Cloud
Server derselben Kategorie von anderen Entwicklern, sortiert nach Sternen.
So liest du eine Karte:★ GitHub-SterneiDrei Säulen: Bewertungen · Checks · NutzungiAktualisiert = neueste Version in der Registry veröffentlichti
Meeting bot and transcripts for Google Meet, Teams and Zoom. Live or after, speakers labelled.
Render markdown into email-safe HTML, lint drafts for deliverability problems, and preview emails.
MCP server for WhatsApp Cloud API (Meta) — messages, media, and templates via Graph API
Unverändert aus dem Repo des Entwicklers übernommen. PopNx MCP hat den Inhalt weder geprüft noch empfiehlt es ihn. HTML, Bilder und Skripte wurden entfernt. Links führen von dieser Website weg.
Maintenance mode. This project is stable and receives security fixes. New features aren't planned, but issues and pull requests are still welcome.
Google Workspace for AI agents: Gmail, Calendar, Drive, Sheets, Docs, Slides, Tasks, and Contacts as a curated set of 67 Model Context Protocol tools, built on the official Google Workspace CLI (gws).
The gws CLI had a built-in MCP server that was removed in v0.8.0 because it exposed 200-400 tools — causing context window bloat in MCP clients. This server takes a curated approach: you choose which Google services to expose, and only a focused set of high-value, narrowly scoped operations are registered as tools. Every tool declares all four MCP annotation hints — readOnlyHint, destructiveHint, idempotentHint, openWorldHint — so clients can reason about side effects, know which writes are safe to retry, and surface clearer consent prompts. This is the permissions leg of trust infrastructure for agents: a deliberately narrow tool surface, side effects declared on every tool, and no freestanding send tool — the only outbound email an agent can trigger is calendar invite/update notifications, opt-in via sendUpdates and off by default.
gws CLI installed and authenticated (npm install -g @googleworkspace/cli && gws auth login)This server exposes no freestanding send tool — gmail_drafts_create explicitly does not send, and the only outbound email an agent can trigger is calendar invite/update notifications via sendUpdates, which is enum-validated and defaults to none. The token gws auth login mints is broader than that.
gws auth login opens a scope picker listing nine scopes. The default grant is seven: full read-write drive, spreadsheets, gmail.modify (Google documents it as "Read, compose, and send emails"), calendar, documents, presentations, and tasks — the same seven you get running non-interactively as DEFAULT_SCOPES.
The other two rows are Cloud Pub/Sub and Cloud Platform, and neither is part of the default grant — gws auth login --help describes --full as "Request all scopes incl. pubsub + cloud-platform."
Which rows start *checked* has not been verified against a live picker — the seven above are the documented default grant, not an observation of the TUI. Read the checkboxes before pressing Enter rather than trusting this paragraph.
So the token on disk can send mail and rewrite Drive even though nothing here will. Deselect what you do not need in the picker, or:
gws auth login --readonly # read-only across services
-s gmail limits the picker to Gmail, per the flag's own help text ("Comma-separated service names to limit scope picker"). It cannot pull in cloud-platform or pubsub, because those two are reachable only through --full. It also replaces any credential you already have rather than adding to it. To add a service to an existing login, use gws-mcp-server auth login.
On Linux there is no keyring, and the encryption key is a file next to the data it encrypts. gws enables the keyring crate's native backends only for macOS and Windows; on every other platform the dependency is declared with no backend feature, so the store falls through to writing .encryption_key into ~/.config/gws/. That file is not a backup of a key held elsewhere — it is the key, and the credential store's own doc comment says it is never deleted. Setting GOOGLE_WORKSPACE_CLI_KEYRING_BACKEND=file changes nothing there because that is already the only path. On macOS and Windows the key file is removed once the OS keyring holds the key. If you run this headless on Linux, treat `~/.config/gws/` as a password file: anyone who can read the directory has the credentials.
The people_* tools need the contacts scope, which the default gws auth login (and the seven-scope grant described above) does not request.
Add Contacts with `gws-mcp-server auth login -s people`, not `gws auth login -s people`. gws's own filtered login replaces the saved credential with only the filtered scopes, so it drops Drive, Gmail, Calendar and everything else you had granted (#68, tracked upstream in googleworkspace/cli#673). This server's auth login does incremental authorization instead:
gws-mcp-server auth login -s people # keep existing scopes, add contacts
gws-mcp-server auth login -s people --dry-run # print the gws command first
It reads the scopes the saved credential holds from gws auth status, adds the requested services' scopes, and runs gws auth login --scopes . You consent once in the browser, and afterwards it re-reads gws auth status and exits non-zero if any scope you asked for (old or new) wasn't granted, for example because a box was unticked on Google's consent screen. gws saves whatever the consent screen granted, so an unticked box does remove that scope from the saved credential. Before the browser opens, and again if anything is missing afterwards, the command prints the gws auth login --scopes "..." command that re-requests the complete list. It needs a client configured for gws auth login already, and it exits without changing anything if the credential already covers the services.
| Flag | Effect |
|---|---|
--services, -s | Services whose scopes to add (required) |
--readonly | Add each service's read-only scope (skipped when the read-write scope is already held) |
--dry-run | Print the gws auth login command without running it |
--replace | The old behaviour: replace the credential with only the requested scopes |
--gws-path | gws binary, same as the server flag |
gws auth status discovers scopes live (refresh-token exchange plus Google's token-info lookup), so the scopes field can be missing when that fails. When a saved credential exists but its scopes can't be read, auth login refuses rather than guessing; fix the status lookup and retry. If you really mean to start over with only the requested scopes, pass --replace. With no saved credential at all it is a first login and requests just the services you name.
If you prefer to run gws directly, pass the complete list yourself. For a credential that used exactly the default seven scopes described above:
gws auth login --scopes "https://www.googleapis.com/auth/drive,https://www.googleapis.com/auth/spreadsheets,https://www.googleapis.com/auth/gmail.modify,https://www.googleapis.com/auth/calendar,https://www.googleapis.com/auth/documents,https://www.googleapis.com/auth/presentations,https://www.googleapis.com/auth/tasks,https://www.googleapis.com/auth/contacts"
The People API must also be enabled on your own GCP project — a one-time step separate from OAuth, since a new scope grant doesn't enable a new API by itself:
gcloud services enable people.googleapis.com --project=<your-project-id>
Without both steps, people_* tools fail: a missing scope surfaces as a 401/403 from Google, and a disabled API surfaces as a distinct 403 naming the API and a console link to enable it.
# Install
npm install -g gws-mcp-server
git clone https://github.com/conorbronsdon/gws-mcp-server.git cd gws-mcp-server npm install && npm run build
.mcp.json){
"mcpServers": { "google-workspace": { "command": "npx", "args": [ "gws-mcp-server", "--services", "drive,sheets,calendar,docs,slides,gmail,tasks" ] } } }
Contacts is opt-in because it needs additional authentication and API setup. After completing the Contacts prerequisites, append people to the service list:
"args": ["gws-mcp-server", "--services", "drive,calendar,people"]
claude_desktop_config.json){
"mcpServers": { "google-workspace": { "command": "npx", "args": [ "gws-mcp-server", "--services", "drive,sheets,calendar" ] } } }
| Flag | Description | Default |
|---|---|---|
--services, -s | Comma-separated list of services to expose | Core services; people is opt-in |
--gws-path | Path to the gws binary | GWS_BINARY or gws |
--read-only | Register only the read-only tools | off |
On Windows, npm's gws.cmd is resolved to its JavaScript entry point and run with Node. --gws-path or GWS_BINARY may also point directly to a JavaScript entry point or .exe.
--read-only--read-only registers 26 tools instead of the default 61. Every tool that writes to Google is left unregistered, so it never appears in tools/list and there is nothing for an agent to call — including gmail_drafts_create, which is a write even though it never sends. drive_files_download stays, since it reads. Explicitly adding people raises those counts to 29 read-only tools out of 67 total.
gws-mcp-server --read-only
gws-mcp-server --read-only --services drive,calendar # combines with -s
This constrains the agent, not the credential. The token on disk keeps whatever scopes it was granted, and anything else on the machine can still use it. gws auth login --readonly is what narrows the token; the two are complementary. For an MCP server the agent is the threat model, but that is the limit of the claim.
Every registered tool rides along in each conversation. The default registry is roughly 46 KB of tools/list payload (~11.8K tokens); opting into people raises the full registry to roughly 52 KB (~13.2K tokens). The two flags above compose, and dropping whole services you don't use is the cheapest context win there is:
gws-mcp-server --services calendar # calendar assistant: 6 tools
gws-mcp-server --services drive,docs # document work, nothing else gws-mcp-server --read-only --services drive,docs,sheets # research setup: reads only
In .mcp.json or claude_desktop_config.json, the same trimming is just editing the args array:
"args": ["gws-mcp-server", "--services", "drive,calendar"]
A service's tool count (headers below) tracks its context cost: dropping drive (24 tools) saves the most, docs (3 tools) the least. There is no per-tool exclude flag today — if service granularity is too coarse for your setup, open an issue describing the split you need.
#
Nur der Anfang wird gezeigt; den Rest findest du im Repo