Gmail MCP Server: What It Does and How to Connect Gmail to AI in 2026
Google has built an official Gmail MCP server — but it’s not something you can just switch on. It’s called the Gmail MCP server (gmailmcp.googleapis.com), and it’s real, Google-hosted, and documented on Google’s own developer site. The catch: it’s in Developer Preview, not general availability, and using it means creating a Google Cloud project, enabling the Gmail API, and generating your own OAuth client — a setup built for developers, not a “click authorize” flow like Gmail users get with, say, a Chrome extension.
The other thing worth knowing before you go looking for it: even once connected, an MCP server hands Gmail to an AI inside a conversation you start. It’s a doorway, not a worker. Nothing watches your inbox for you, nothing fires when an email arrives, and nothing runs while the chat is closed. Here’s exactly what the Gmail MCP does, how to actually get it running, where it stops — and what to use when you want Gmail work that runs on its own.
What the Gmail MCP server does
Model Context Protocol (MCP) is the open standard that lets an AI client — Claude, ChatGPT, Google Antigravity, and others — talk to an outside app through a shared interface. Google’s own Gmail MCP server, announced alongside Workspace MCP servers for Drive, Calendar, People, and Chat at Google Cloud Next ‘26, exposes a specific set of tools:
- search_threads and get_thread — find and read email conversations.
- list_labels and create_label — see and create the labels in your inbox.
- label_message / unlabel_message and label_thread / unlabel_thread — organize mail by applying or removing labels.
- create_draft and list_drafts — write an unsent message and see what’s queued.
Notice what’s missing: there’s no send_message tool. Google’s own Gmail MCP server can draft, search, read, and label — it cannot send. If you want an AI to actually transmit an email through this server, it can’t; a human has to open the draft and hit send. That’s consistent with where Gmail’s AI features generally sit today — drafting and summarizing, not autonomous sending.
For most people, this is genuinely useful for ad-hoc work inside a chat: “find my last thread with Acme and draft a reply,” “label everything from this sender as Receipts,” “what’s still sitting in my drafts folder.” It’s not useful for anything that needs to happen automatically.
How to connect Gmail to AI in 2026
Two real paths exist right now, and neither is a one-click toggle:
The official path (developer-oriented, preview-gated). Google’s docs require: (1) a Google Cloud project, (2) enabling the Gmail API and its MCP service variant in that project, (3) creating an OAuth 2.0 client ID and secret with the right redirect URI for your AI client, and (4) enrollment in the Google Workspace Developer Preview Program, since the server isn’t generally available yet. Once that’s done, you point an MCP-compatible client at gmailmcp.googleapis.com and authorize it against your Gmail account.
The community path (available today, still requires your own credentials). Independent developers have published Gmail MCP servers — GongRzhe’s Gmail-MCP-Server for Claude Desktop and Taylor Wilsdon’s google_workspace_mcp among the more actively maintained — that wrap the same underlying Gmail API. These aren’t gated behind a preview program, but Google still requires you to create your own OAuth client in a Google Cloud project to use them, because that’s how the Gmail API grants access regardless of who wrote the MCP wrapper.
Either way, “connect Gmail to AI” today means touching the Google Cloud console. There’s no version of this yet where a non-technical person just flips a switch.
Where the Gmail MCP stops
None of this is a knock on MCP — it’s the shape of the protocol, and Google’s preview-gating makes the gap even more visible. Four limits show up the moment you want more than a conversation:
- It only works inside a chat you start. Close the window and nothing happens. The AI doesn’t watch your inbox; it waits for you to ask.
- No triggers. A new email from a client, a message matching a label, an invoice landing in your inbox — none of these can start anything through MCP. There’s no “when this lands in Gmail, do that.”
- It’s one app at a time. The Gmail MCP knows Gmail. Getting an attachment into Drive, a lead into your CRM, and a summary into Slack means wiring up a separate MCP server for each and hoping your client can juggle them in one turn.
- You own the plumbing and the scopes — and with Gmail specifically, that plumbing currently includes a Google Cloud project, an OAuth client, and possibly a preview-program waitlist.
So even in the best case, the Gmail MCP is a way to ask about your inbox and draft replies inside a chat. It’s not a way to make Gmail run — to have work happen on a schedule or in reaction to a new email, across the other tools that email touches.
Running Gmail work that doesn’t need a chat open
That “run on its own, across apps” gap is exactly where Carly fits. Carly connects to Gmail natively — no Cloud project, no OAuth client to generate, no preview program to join — and to the ~260 other apps it supports, plus anything with a public API through your own key. The difference from MCP is the important part: Carly’s workflows are triggered and scheduled, so Gmail work happens whether or not anyone has a chat window open.
A few things MCP can’t do but a Carly workflow can:
- When an email arrives from a client → pull the relevant thread history, draft a reply, and route it to the right person for approval — automatically, the moment it lands.
- Every morning → summarize unread emails older than 24 hours and send the digest to Slack.
- When an invoice hits the inbox → save the attachment to Drive, log the amount in a spreadsheet, and label the email “Processed.”
The non-AI steps — the moving, matching, and routing between apps — are free and unlimited, the Zapier-style backbone of the workflow. The AI steps (drafting, summarizing, deciding) start at $35/month. You describe the outcome in plain language and Carly wires up the Gmail connection and everything downstream.
If you want to interrogate your inbox from inside a chat and don’t mind setting up a Google Cloud project, Google’s own Gmail MCP server (once you’re off the waitlist) does that. If you want Gmail to actually do things — on a trigger, on a schedule, across every app an email touches — that’s the job MCP wasn’t built for, and it’s the one Carly was.
FAQ
Does Gmail have an official MCP server?
Yes — Google built and hosts one at gmailmcp.googleapis.com, documented on Google’s own developer site. It’s currently in Developer Preview, not general availability, and requires a Google Cloud project plus your own OAuth client to use.
Can the Gmail MCP server send emails? No. Google’s official Gmail MCP server can search, read, label, and create drafts — but there’s no send tool. A human has to open the draft and send it themselves.
Can the Gmail MCP trigger automations? No. MCP is request/response inside an AI chat — it has no triggers and nothing runs when the conversation is closed. For event- or schedule-driven Gmail work across apps, you need a workflow tool like Carly rather than an MCP server.
Can I connect Gmail to AI without coding or hosting a server? Yes. You don’t have to touch the Gmail API, a Google Cloud project, or any OAuth client. Carly connects to Gmail for you and lets you build the automation in plain language — describe what you want to happen and it wires up Gmail and the other apps involved, with no server to host and no code to write.
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
